Tag Archives: conversion

Analyze Apache Parquet optimized data using Amazon Kinesis Data Firehose, Amazon Athena, and Amazon Redshift

Post Syndicated from Roy Hasson original https://aws.amazon.com/blogs/big-data/analyzing-apache-parquet-optimized-data-using-amazon-kinesis-data-firehose-amazon-athena-and-amazon-redshift/

Amazon Kinesis Data Firehose is the easiest way to capture and stream data into a data lake built on Amazon S3. This data can be anything—from AWS service logs like AWS CloudTrail log files, Amazon VPC Flow Logs, Application Load Balancer logs, and others. It can also be IoT events, game events, and much more. To efficiently query this data, a time-consuming ETL (extract, transform, and load) process is required to massage and convert the data to an optimal file format, which increases the time to insight. This situation is less than ideal, especially for real-time data that loses its value over time.

To solve this common challenge, Kinesis Data Firehose can now save data to Amazon S3 in Apache Parquet or Apache ORC format. These are optimized columnar formats that are highly recommended for best performance and cost-savings when querying data in S3. This feature directly benefits you if you use Amazon Athena, Amazon Redshift, AWS Glue, Amazon EMR, or any other big data tools that are available from the AWS Partner Network and through the open-source community.

Amazon Connect is a simple-to-use, cloud-based contact center service that makes it easy for any business to provide a great customer experience at a lower cost than common alternatives. Its open platform design enables easy integration with other systems. One of those systems is Amazon Kinesis—in particular, Kinesis Data Streams and Kinesis Data Firehose.

What’s really exciting is that you can now save events from Amazon Connect to S3 in Apache Parquet format. You can then perform analytics using Amazon Athena and Amazon Redshift Spectrum in real time, taking advantage of this key performance and cost optimization. Of course, Amazon Connect is only one example. This new capability opens the door for a great deal of opportunity, especially as organizations continue to build their data lakes.

Amazon Connect includes an array of analytics views in the Administrator dashboard. But you might want to run other types of analysis. In this post, I describe how to set up a data stream from Amazon Connect through Kinesis Data Streams and Kinesis Data Firehose and out to S3, and then perform analytics using Athena and Amazon Redshift Spectrum. I focus primarily on the Kinesis Data Firehose support for Parquet and its integration with the AWS Glue Data Catalog, Amazon Athena, and Amazon Redshift.

Solution overview

Here is how the solution is laid out:

 

 

The following sections walk you through each of these steps to set up the pipeline.

1. Define the schema

When Kinesis Data Firehose processes incoming events and converts the data to Parquet, it needs to know which schema to apply. The reason is that many times, incoming events contain all or some of the expected fields based on which values the producers are advertising. A typical process is to normalize the schema during a batch ETL job so that you end up with a consistent schema that can easily be understood and queried. Doing this introduces latency due to the nature of the batch process. To overcome this issue, Kinesis Data Firehose requires the schema to be defined in advance.

To see the available columns and structures, see Amazon Connect Agent Event Streams. For the purpose of simplicity, I opted to make all the columns of type String rather than create the nested structures. But you can definitely do that if you want.

The simplest way to define the schema is to create a table in the Amazon Athena console. Open the Athena console, and paste the following create table statement, substituting your own S3 bucket and prefix for where your event data will be stored. A Data Catalog database is a logical container that holds the different tables that you can create. The default database name shown here should already exist. If it doesn’t, you can create it or use another database that you’ve already created.

CREATE EXTERNAL TABLE default.kfhconnectblog (
  awsaccountid string,
  agentarn string,
  currentagentsnapshot string,
  eventid string,
  eventtimestamp string,
  eventtype string,
  instancearn string,
  previousagentsnapshot string,
  version string
)
STORED AS parquet
LOCATION 's3://your_bucket/kfhconnectblog/'
TBLPROPERTIES ("parquet.compression"="SNAPPY")

That’s all you have to do to prepare the schema for Kinesis Data Firehose.

2. Define the data streams

Next, you need to define the Kinesis data streams that will be used to stream the Amazon Connect events.  Open the Kinesis Data Streams console and create two streams.  You can configure them with only one shard each because you don’t have a lot of data right now.

3. Define the Kinesis Data Firehose delivery stream for Parquet

Let’s configure the Data Firehose delivery stream using the data stream as the source and Amazon S3 as the output. Start by opening the Kinesis Data Firehose console and creating a new data delivery stream. Give it a name, and associate it with the Kinesis data stream that you created in Step 2.

As shown in the following screenshot, enable Record format conversion (1) and choose Apache Parquet (2). As you can see, Apache ORC is also supported. Scroll down and provide the AWS Glue Data Catalog database name (3) and table names (4) that you created in Step 1. Choose Next.

To make things easier, the output S3 bucket and prefix fields are automatically populated using the values that you defined in the LOCATION parameter of the create table statement from Step 1. Pretty cool. Additionally, you have the option to save the raw events into another location as defined in the Source record S3 backup section. Don’t forget to add a trailing forward slash “ / “ so that Data Firehose creates the date partitions inside that prefix.

On the next page, in the S3 buffer conditions section, there is a note about configuring a large buffer size. The Parquet file format is highly efficient in how it stores and compresses data. Increasing the buffer size allows you to pack more rows into each output file, which is preferred and gives you the most benefit from Parquet.

Compression using Snappy is automatically enabled for both Parquet and ORC. You can modify the compression algorithm by using the Kinesis Data Firehose API and update the OutputFormatConfiguration.

Be sure to also enable Amazon CloudWatch Logs so that you can debug any issues that you might run into.

Lastly, finalize the creation of the Firehose delivery stream, and continue on to the next section.

4. Set up the Amazon Connect contact center

After setting up the Kinesis pipeline, you now need to set up a simple contact center in Amazon Connect. The Getting Started page provides clear instructions on how to set up your environment, acquire a phone number, and create an agent to accept calls.

After setting up the contact center, in the Amazon Connect console, choose your Instance Alias, and then choose Data Streaming. Under Agent Event, choose the Kinesis data stream that you created in Step 2, and then choose Save.

At this point, your pipeline is complete.  Agent events from Amazon Connect are generated as agents go about their day. Events are sent via Kinesis Data Streams to Kinesis Data Firehose, which converts the event data from JSON to Parquet and stores it in S3. Athena and Amazon Redshift Spectrum can simply query the data without any additional work.

So let’s generate some data. Go back into the Administrator console for your Amazon Connect contact center, and create an agent to handle incoming calls. In this example, I creatively named mine Agent One. After it is created, Agent One can get to work and log into their console and set their availability to Available so that they are ready to receive calls.

To make the data a bit more interesting, I also created a second agent, Agent Two. I then made some incoming and outgoing calls and caused some failures to occur, so I now have enough data available to analyze.

5. Analyze the data with Athena

Let’s open the Athena console and run some queries. One thing you’ll notice is that when we created the schema for the dataset, we defined some of the fields as Strings even though in the documentation they were complex structures.  The reason for doing that was simply to show some of the flexibility of Athena to be able to parse JSON data. However, you can define nested structures in your table schema so that Kinesis Data Firehose applies the appropriate schema to the Parquet file.

Let’s run the first query to see which agents have logged into the system.

The query might look complex, but it’s fairly straightforward:

WITH dataset AS (
  SELECT 
    from_iso8601_timestamp(eventtimestamp) AS event_ts,
    eventtype,
    -- CURRENT STATE
    json_extract_scalar(
      currentagentsnapshot,
      '$.agentstatus.name') AS current_status,
    from_iso8601_timestamp(
      json_extract_scalar(
        currentagentsnapshot,
        '$.agentstatus.starttimestamp')) AS current_starttimestamp,
    json_extract_scalar(
      currentagentsnapshot, 
      '$.configuration.firstname') AS current_firstname,
    json_extract_scalar(
      currentagentsnapshot,
      '$.configuration.lastname') AS current_lastname,
    json_extract_scalar(
      currentagentsnapshot, 
      '$.configuration.username') AS current_username,
    json_extract_scalar(
      currentagentsnapshot, 
      '$.configuration.routingprofile.defaultoutboundqueue.name') AS               current_outboundqueue,
    json_extract_scalar(
      currentagentsnapshot, 
      '$.configuration.routingprofile.inboundqueues[0].name') as current_inboundqueue,
    -- PREVIOUS STATE
    json_extract_scalar(
      previousagentsnapshot, 
      '$.agentstatus.name') as prev_status,
    from_iso8601_timestamp(
      json_extract_scalar(
        previousagentsnapshot, 
       '$.agentstatus.starttimestamp')) as prev_starttimestamp,
    json_extract_scalar(
      previousagentsnapshot, 
      '$.configuration.firstname') as prev_firstname,
    json_extract_scalar(
      previousagentsnapshot, 
      '$.configuration.lastname') as prev_lastname,
    json_extract_scalar(
      previousagentsnapshot, 
      '$.configuration.username') as prev_username,
    json_extract_scalar(
      previousagentsnapshot, 
      '$.configuration.routingprofile.defaultoutboundqueue.name') as current_outboundqueue,
    json_extract_scalar(
      previousagentsnapshot, 
      '$.configuration.routingprofile.inboundqueues[0].name') as prev_inboundqueue
  from kfhconnectblog
  where eventtype <> 'HEART_BEAT'
)
SELECT
  current_status as status,
  current_username as username,
  event_ts
FROM dataset
WHERE eventtype = 'LOGIN' AND current_username <> ''
ORDER BY event_ts DESC

The query output looks something like this:

Here is another query that shows the sessions each of the agents engaged with. It tells us where they were incoming or outgoing, if they were completed, and where there were missed or failed calls.

WITH src AS (
  SELECT
     eventid,
     json_extract_scalar(currentagentsnapshot, '$.configuration.username') as username,
     cast(json_extract(currentagentsnapshot, '$.contacts') AS ARRAY(JSON)) as c,
     cast(json_extract(previousagentsnapshot, '$.contacts') AS ARRAY(JSON)) as p
  from kfhconnectblog
),
src2 AS (
  SELECT *
  FROM src CROSS JOIN UNNEST (c, p) AS contacts(c_item, p_item)
),
dataset AS (
SELECT 
  eventid,
  username,
  json_extract_scalar(c_item, '$.contactid') as c_contactid,
  json_extract_scalar(c_item, '$.channel') as c_channel,
  json_extract_scalar(c_item, '$.initiationmethod') as c_direction,
  json_extract_scalar(c_item, '$.queue.name') as c_queue,
  json_extract_scalar(c_item, '$.state') as c_state,
  from_iso8601_timestamp(json_extract_scalar(c_item, '$.statestarttimestamp')) as c_ts,
  
  json_extract_scalar(p_item, '$.contactid') as p_contactid,
  json_extract_scalar(p_item, '$.channel') as p_channel,
  json_extract_scalar(p_item, '$.initiationmethod') as p_direction,
  json_extract_scalar(p_item, '$.queue.name') as p_queue,
  json_extract_scalar(p_item, '$.state') as p_state,
  from_iso8601_timestamp(json_extract_scalar(p_item, '$.statestarttimestamp')) as p_ts
FROM src2
)
SELECT 
  username,
  c_channel as channel,
  c_direction as direction,
  p_state as prev_state,
  c_state as current_state,
  c_ts as current_ts,
  c_contactid as id
FROM dataset
WHERE c_contactid = p_contactid
ORDER BY id DESC, current_ts ASC

The query output looks similar to the following:

6. Analyze the data with Amazon Redshift Spectrum

With Amazon Redshift Spectrum, you can query data directly in S3 using your existing Amazon Redshift data warehouse cluster. Because the data is already in Parquet format, Redshift Spectrum gets the same great benefits that Athena does.

Here is a simple query to show querying the same data from Amazon Redshift. Note that to do this, you need to first create an external schema in Amazon Redshift that points to the AWS Glue Data Catalog.

SELECT 
  eventtype,
  json_extract_path_text(currentagentsnapshot,'agentstatus','name') AS current_status,
  json_extract_path_text(currentagentsnapshot, 'configuration','firstname') AS current_firstname,
  json_extract_path_text(currentagentsnapshot, 'configuration','lastname') AS current_lastname,
  json_extract_path_text(
    currentagentsnapshot,
    'configuration','routingprofile','defaultoutboundqueue','name') AS current_outboundqueue,
FROM default_schema.kfhconnectblog

The following shows the query output:

Summary

In this post, I showed you how to use Kinesis Data Firehose to ingest and convert data to columnar file format, enabling real-time analysis using Athena and Amazon Redshift. This great feature enables a level of optimization in both cost and performance that you need when storing and analyzing large amounts of data. This feature is equally important if you are investing in building data lakes on AWS.

 


Additional Reading

If you found this post useful, be sure to check out Analyzing VPC Flow Logs with Amazon Kinesis Firehose, Amazon Athena, and Amazon QuickSight and Work with partitioned data in AWS Glue.


About the Author

Roy Hasson is a Global Business Development Manager for AWS Analytics. He works with customers around the globe to design solutions to meet their data processing, analytics and business intelligence needs. Roy is big Manchester United fan cheering his team on and hanging out with his family.

 

 

 

Implement continuous integration and delivery of serverless AWS Glue ETL applications using AWS Developer Tools

Post Syndicated from Prasad Alle original https://aws.amazon.com/blogs/big-data/implement-continuous-integration-and-delivery-of-serverless-aws-glue-etl-applications-using-aws-developer-tools/

AWS Glue is an increasingly popular way to develop serverless ETL (extract, transform, and load) applications for big data and data lake workloads. Organizations that transform their ETL applications to cloud-based, serverless ETL architectures need a seamless, end-to-end continuous integration and continuous delivery (CI/CD) pipeline: from source code, to build, to deployment, to product delivery. Having a good CI/CD pipeline can help your organization discover bugs before they reach production and deliver updates more frequently. It can also help developers write quality code and automate the ETL job release management process, mitigate risk, and more.

AWS Glue is a fully managed data catalog and ETL service. It simplifies and automates the difficult and time-consuming tasks of data discovery, conversion, and job scheduling. AWS Glue crawls your data sources and constructs a data catalog using pre-built classifiers for popular data formats and data types, including CSV, Apache Parquet, JSON, and more.

When you are developing ETL applications using AWS Glue, you might come across some of the following CI/CD challenges:

  • Iterative development with unit tests
  • Continuous integration and build
  • Pushing the ETL pipeline to a test environment
  • Pushing the ETL pipeline to a production environment
  • Testing ETL applications using real data (live test)
  • Exploring and validating data

In this post, I walk you through a solution that implements a CI/CD pipeline for serverless AWS Glue ETL applications supported by AWS Developer Tools (including AWS CodePipeline, AWS CodeCommit, and AWS CodeBuild) and AWS CloudFormation.

Solution overview

The following diagram shows the pipeline workflow:

This solution uses AWS CodePipeline, which lets you orchestrate and automate the test and deploy stages for ETL application source code. The solution consists of a pipeline that contains the following stages:

1.) Source Control: In this stage, the AWS Glue ETL job source code and the AWS CloudFormation template file for deploying the ETL jobs are both committed to version control. I chose to use AWS CodeCommit for version control.

To get the ETL job source code and AWS CloudFormation template, download the gluedemoetl.zip file. This solution is developed based on a previous post, Build a Data Lake Foundation with AWS Glue and Amazon S3.

2.) LiveTest: In this stage, all resources—including AWS Glue crawlers, jobs, S3 buckets, roles, and other resources that are required for the solution—are provisioned, deployed, live tested, and cleaned up.

The LiveTest stage includes the following actions:

  • Deploy: In this action, all the resources that are required for this solution (crawlers, jobs, buckets, roles, and so on) are provisioned and deployed using an AWS CloudFormation template.
  • AutomatedLiveTest: In this action, all the AWS Glue crawlers and jobs are executed and data exploration and validation tests are performed. These validation tests include, but are not limited to, record counts in both raw tables and transformed tables in the data lake and any other business validations. I used AWS CodeBuild for this action.
  • LiveTestApproval: This action is included for the cases in which a pipeline administrator approval is required to deploy/promote the ETL applications to the next stage. The pipeline pauses in this action until an administrator manually approves the release.
  • LiveTestCleanup: In this action, all the LiveTest stage resources, including test crawlers, jobs, roles, and so on, are deleted using the AWS CloudFormation template. This action helps minimize cost by ensuring that the test resources exist only for the duration of the AutomatedLiveTest and LiveTestApproval

3.) DeployToProduction: In this stage, all the resources are deployed using the AWS CloudFormation template to the production environment.

Try it out

This code pipeline takes approximately 20 minutes to complete the LiveTest test stage (up to the LiveTest approval stage, in which manual approval is required).

To get started with this solution, choose Launch Stack:

This creates the CI/CD pipeline with all of its stages, as described earlier. It performs an initial commit of the sample AWS Glue ETL job source code to trigger the first release change.

In the AWS CloudFormation console, choose Create. After the template finishes creating resources, you see the pipeline name on the stack Outputs tab.

After that, open the CodePipeline console and select the newly created pipeline. Initially, your pipeline’s CodeCommit stage shows that the source action failed.

Allow a few minutes for your new pipeline to detect the initial commit applied by the CloudFormation stack creation. As soon as the commit is detected, your pipeline starts. You will see the successful stage completion status as soon as the CodeCommit source stage runs.

In the CodeCommit console, choose Code in the navigation pane to view the solution files.

Next, you can watch how the pipeline goes through the LiveTest stage of the deploy and AutomatedLiveTest actions, until it finally reaches the LiveTestApproval action.

At this point, if you check the AWS CloudFormation console, you can see that a new template has been deployed as part of the LiveTest deploy action.

At this point, make sure that the AWS Glue crawlers and the AWS Glue job ran successfully. Also check whether the corresponding databases and external tables have been created in the AWS Glue Data Catalog. Then verify that the data is validated using Amazon Athena, as shown following.

Open the AWS Glue console, and choose Databases in the navigation pane. You will see the following databases in the Data Catalog:

Open the Amazon Athena console, and run the following queries. Verify that the record counts are matching.

SELECT count(*) FROM "nycitytaxi_gluedemocicdtest"."data";
SELECT count(*) FROM "nytaxiparquet_gluedemocicdtest"."datalake";

The following shows the raw data:

The following shows the transformed data:

The pipeline pauses the action until the release is approved. After validating the data, manually approve the revision on the LiveTestApproval action on the CodePipeline console.

Add comments as needed, and choose Approve.

The LiveTestApproval stage now appears as Approved on the console.

After the revision is approved, the pipeline proceeds to use the AWS CloudFormation template to destroy the resources that were deployed in the LiveTest deploy action. This helps reduce cost and ensures a clean test environment on every deployment.

Production deployment is the final stage. In this stage, all the resources—AWS Glue crawlers, AWS Glue jobs, Amazon S3 buckets, roles, and so on—are provisioned and deployed to the production environment using the AWS CloudFormation template.

After successfully running the whole pipeline, feel free to experiment with it by changing the source code stored on AWS CodeCommit. For example, if you modify the AWS Glue ETL job to generate an error, it should make the AutomatedLiveTest action fail. Or if you change the AWS CloudFormation template to make its creation fail, it should affect the LiveTest deploy action. The objective of the pipeline is to guarantee that all changes that are deployed to production are guaranteed to work as expected.

Conclusion

In this post, you learned how easy it is to implement CI/CD for serverless AWS Glue ETL solutions with AWS developer tools like AWS CodePipeline and AWS CodeBuild at scale. Implementing such solutions can help you accelerate ETL development and testing at your organization.

If you have questions or suggestions, please comment below.

 


Additional Reading

If you found this post useful, be sure to check out Implement Continuous Integration and Delivery of Apache Spark Applications using AWS and Build a Data Lake Foundation with AWS Glue and Amazon S3.

 


About the Authors

Prasad Alle is a Senior Big Data Consultant with AWS Professional Services. He spends his time leading and building scalable, reliable Big data, Machine learning, Artificial Intelligence and IoT solutions for AWS Enterprise and Strategic customers. His interests extend to various technologies such as Advanced Edge Computing, Machine learning at Edge. In his spare time, he enjoys spending time with his family.

 
Luis Caro is a Big Data Consultant for AWS Professional Services. He works with our customers to provide guidance and technical assistance on big data projects, helping them improving the value of their solutions when using AWS.

 

 

 

More power to your Pi

Post Syndicated from James Adams original https://www.raspberrypi.org/blog/pi-power-supply-chip/

It’s been just over three weeks since we launched the new Raspberry Pi 3 Model B+. Although the product is branded Raspberry Pi 3B+ and not Raspberry Pi 4, a serious amount of engineering was involved in creating it. The wireless networking, USB/Ethernet hub, on-board power supplies, and BCM2837 chip were all upgraded: together these represent almost all the circuitry on the board! Today, I’d like to tell you about the work that has gone into creating a custom power supply chip for our newest computer.

Raspberry Pi 3 Model B+, with custome power supply chip

The new Raspberry Pi 3B+, sporting a new, custom power supply chip (bottom left-hand corner)

Successful launch

The Raspberry Pi 3B+ has been well received, and we’ve enjoyed hearing feedback from the community as well as reading the various reviews and articles highlighting the solid improvements in wireless networking, Ethernet, CPU, and thermal performance of the new board. Gareth Halfacree’s post here has some particularly nice graphs showing the increased performance as well as how the Pi 3B+ keeps cool under load due to the new CPU package that incorporates a metal heat spreader. The Raspberry Pi production lines at the Sony UK Technology Centre are running at full speed, and it seems most people who want to get hold of the new board are able to find one in stock.

Powering your Pi

One of the most critical but often under-appreciated elements of any electronic product, particularly one such as Raspberry Pi with lots of complex on-board silicon (processor, networking, high-speed memory), is the power supply. In fact, the Raspberry Pi 3B+ has no fewer than six different voltage rails: two at 3.3V — one special ‘quiet’ one for audio, and one for everything else; 1.8V; 1.2V for the LPDDR2 memory; and 1.2V nominal for the CPU core. Note that the CPU voltage is actually raised and lowered on the fly as the speed of the CPU is increased and decreased depending on how hard the it is working. The sixth rail is 5V, which is the master supply that all the others are created from, and the output voltage for the four downstream USB ports; this is what the mains power adaptor is supplying through the micro USB power connector.

Power supply primer

There are two common classes of power supply circuits: linear regulators and switching regulators. Linear regulators work by creating a lower, regulated voltage from a higher one. In simple terms, they monitor the output voltage against an internally generated reference and continually change their own resistance to keep the output voltage constant. Switching regulators work in a different way: they ‘pump’ energy by first storing the energy coming from the source supply in a reactive component (usually an inductor, sometimes a capacitor) and then releasing it to the regulated output supply. The switches in switching regulators effect this energy transfer by first connecting the inductor (or capacitor) to store the source energy, and then switching the circuit so the energy is released to its destination.

Linear regulators produce smoother, less noisy output voltages, but they can only convert to a lower voltage, and have to dissipate energy to do so. The higher the output current and the voltage difference across them is, the more energy is lost as heat. On the other hand, switching supplies can, depending on their design, convert any voltage to any other voltage and can be much more efficient (efficiencies of 90% and above are not uncommon). However, they are more complex and generate noisier output voltages.

Designers use both types of regulators depending on the needs of the downstream circuit: for low-voltage drops, low current, or low noise, linear regulators are usually the right choice, while switching regulators are used for higher power or when efficiency of conversion is required. One of the simplest switching-mode power supply circuits is the buck converter, used to create a lower voltage from a higher one, and this is what we use on the Pi.

A history lesson

The BCM2835 processor chip (found on the original Raspberry Pi Model B and B+, as well as on the Zero products) has on-chip power supplies: one switch-mode regulator for the core voltage, as well as a linear one for the LPDDR2 memory supply. This meant that in addition to 5V, we only had to provide 3.3V and 1.8V on the board, which was relatively simple to do using cheap, off-the-shelf parts.

Pi Zero sporting a BCM2835 processor which only needs 2 external switchers (the components clustered behind the camera port)

When we moved to the BCM2836 for Raspberry Pi Model 2 (and subsequently to the BCM2837A1 and B0 for Raspberry Pi 3B and 3B+), the core supply and the on-chip LPDDR2 memory supply were not up to the job of supplying the extra processor cores and larger memory, so we removed them. (We also used the recovered chip area to help fit in the new quad-core ARM processors.) The upshot of this was that we had to supply these power rails externally for the Raspberry Pi 2 and models thereafter. Moreover, we also had to provide circuitry to sequence them correctly in order to control exactly when they power up compared to the other supplies on the board.

Power supply design is tricky (but critical)

Raspberry Pi boards take in 5V from the micro USB socket and have to generate the other required supplies from this. When 5V is first connected, each of these other supplies must ‘start up’, meaning go from ‘off’, or 0V, to their correct voltage in some short period of time. The order of the supplies starting up is often important: commonly, there are structures inside a chip that form diodes between supply rails, and bringing supplies up in the wrong order can sometimes ‘turn on’ these diodes, causing them to conduct, with undesirable consequences. Silicon chips come with a data sheet specifying what supplies (voltages and currents) are needed and whether they need to be low-noise, in what order they must power up (and in some cases down), and sometimes even the rate at which the voltages must power up and down.

A Pi3. Power supply components are clustered bottom left next to the micro USB, middle (above LPDDR2 chip which is on the bottom of the PCB) and above the A/V jack.

In designing the power chain for the Pi 2 and 3, the sequencing was fairly straightforward: power rails power up in order of voltage (5V, 3.3V, 1.8V, 1.2V). However, the supplies were all generated with individual, discrete devices. Therefore, I spent quite a lot of time designing circuitry to control the sequencing — even with some design tricks to reduce component count, quite a few sequencing components are required. More complex systems generally use a Power Management Integrated Circuit (PMIC) with multiple supplies on a single chip, and many different PMIC variants are made by various manufacturers. Since Raspberry Pi 2 days, I was looking for a suitable PMIC to simplify the Pi design, but invariably (and somewhat counter-intuitively) these were always too expensive compared to my discrete solution, usually because they came with more features than needed.

One device to rule them all

It was way back in May 2015 when I first chatted to Peter Coyle of Exar (Exar were bought by MaxLinear in 2017) about power supply products for Raspberry Pi. We didn’t find a product match then, but in June 2016 Peter, along with Tuomas Hollman and Trevor Latham, visited to pitch the possibility of building a custom power management solution for us.

I was initially sceptical that it could be made cheap enough. However, our discussion indicated that if we could tailor the solution to just what we needed, it could be cost-effective. Over the coming weeks and months, we honed a specification we agreed on from the initial sketches we’d made, and Exar thought they could build it for us at the target price.

The chip we designed would contain all the key supplies required for the Pi on one small device in a cheap QFN package, and it would also perform the required sequencing and voltage monitoring. Moreover, the chip would be flexible to allow adjustment of supply voltages from their default values via I2C; the largest supply would be capable of being adjusted quickly to perform the dynamic core voltage changes needed in order to reduce voltage to the processor when it is idling (to save power), and to boost voltage to the processor when running at maximum speed (1.4 GHz). The supplies on the chip would all be generously specified and could deliver significantly more power than those used on the Raspberry Pi 3. All in all, the chip would contain four switching-mode converters and one low-current linear regulator, this last one being low-noise for the audio circuitry.

The MXL7704 chip

The project was a great success: MaxLinear delivered working samples of first silicon at the end of May 2017 (almost exactly a year after we had kicked off the project), and followed through with production quantities in December 2017 in time for the Raspberry Pi 3B+ production ramp.

The team behind the power supply chip on the Raspberry Pi 3 Model B+ (group of six men, two of whom are holding Raspberry Pi boards)

Front row: Roger with the very first Pi 3B+ prototypes and James with a MXL7704 development board hacked to power a Pi 3. Back row left to right: Will Torgerson, Trevor Latham, Peter Coyle, Tuomas Hollman.

The MXL7704 device has been key to reducing Pi board complexity and therefore overall bill of materials cost. Furthermore, by being able to deliver more power when needed, it has also been essential to increasing the speed of the (newly packaged) BCM2837B0 processor on the 3B+ to 1.4GHz. The result is improvements to both the continuous output current to the CPU (from 3A to 4A) and to the transient performance (i.e. the chip has helped to reduce the ‘transient response’, which is the change in supply voltage due to a sudden current spike that occurs when the processor suddenly demands a large current in a few nanoseconds, as modern CPUs tend to do).

With the MXL7704, the power supply circuitry on the 3B+ is now a lot simpler than the Pi 3B design. This new supply also provides the LPDDR2 memory voltage directly from a switching regulator rather than using linear regulators like the Pi 3, thereby improving energy efficiency. This helps to somewhat offset the extra power that the faster Ethernet, wireless networking, and processor consume. A pleasing side effect of using the new chip is the symmetric board layout of the regulators — it’s easy to see the four switching-mode supplies, given away by four similar-looking blobs (three grey and one brownish), which are the inductors.

Close-up of the power supply chip on the Raspberry Pi 3 Model B+

The Pi 3B+ PMIC MXL7704 — pleasingly symmetric

Kudos

It takes a lot of effort to design a new chip from scratch and get it all the way through to production — we are very grateful to the team at MaxLinear for their hard work, dedication, and enthusiasm. We’re also proud to have created something that will not only power Raspberry Pis, but will also be useful for other product designs: it turns out when you have a low-cost and flexible device, it can be used for many things — something we’re fairly familiar with here at Raspberry Pi! For the curious, the product page (including the data sheet) for the MXL7704 chip is here. Particular thanks go to Peter Coyle, Tuomas Hollman, and Trevor Latham, and also to Jon Cronk, who has been our contact in the US and has had to get up early to attend all our conference calls!

The MXL7704 design team celebrating on Pi Day — it takes a lot of people to design a chip!

I hope you liked reading about some of the effort that has gone into creating the new Pi. It’s nice to finally have a chance to tell people about some of the (increasingly complex) technical work that makes building a $35 computer possible — we’re very pleased with the Raspberry Pi 3B+, and we hope you enjoy using it as much as we’ve enjoyed creating it!

The post More power to your Pi appeared first on Raspberry Pi.

If YouTube-Ripping Sites Are Illegal, What About Tools That Do a Similar Job?

Post Syndicated from Andy original https://torrentfreak.com/if-youtube-ripping-sites-are-illegal-what-about-tools-that-do-a-similar-job-180407/

In 2016, the International Federation of the Phonographic Industry published research which claimed that half of 16 to 24-year-olds use stream-ripping tools to copy music from sites like YouTube.

While this might not have surprised those who regularly participate in the activity, IFPI said that volumes had become so vast that stream-ripping had overtaken pirate site music downloads. That was a big statement.

Probably not coincidentally, just two weeks later IFPI, RIAA, and BPI announced legal action against the world’s largest YouTube ripping site, YouTube-MP3.

“YTMP3 rapidly and seamlessly removes the audio tracks contained in videos streamed from YouTube that YTMP3’s users access, converts those audio tracks to an MP3 format, copies and stores them on YTMP3’s servers, and then distributes copies of the MP3 audio files from its servers to its users in the United States, enabling its users to download those MP3 files to their computers, tablets, or smartphones,” the complaint read.

The labels sued YouTube-MP3 for direct infringement, contributory infringement, vicarious infringement, inducing others to infringe, plus circumvention of technological measures on top. The case was big and one that would’ve been intriguing to watch play out in court, but that never happened.

A year later in September 2017, YouTubeMP3 settled out of court. No details were made public but YouTube-MP3 apparently took all the blame and the court was asked to rule in favor of the labels on all counts.

This certainly gave the impression that what YouTube-MP3 did was illegal and a strong message was sent out to other companies thinking of offering a similar service. However, other onlookers clearly saw the labels’ lawsuit as something to be studied and learned from.

One of those was the operator of NotMP3downloader.com, a site that offers Free MP3 Recorder for YouTube, a tool offering similar functionality to YouTube-MP3 while supposedly avoiding the same legal pitfalls.

Part of that involves audio being processed on the user’s machine – not by stream-ripping as such – but by stream-recording. A subtle difference perhaps, but the site’s operator thinks it’s important.

“After examining the claims made by the copyright holders against youtube-mp3.org, we identified that the charges were based on the three main points. [None] of them are applicable to our product,” he told TF this week.

The first point involves YouTube-MP3’s acts of conversion, storage and distribution of content it had previously culled from YouTube. Copies of unlicensed tracks were clearly held on its own servers, a potent direct infringement risk.

“We don’t have any servers to download, convert or store a copyrighted or any other content from YouTube. Therefore, we do not violate any law or prohibition implied in this part,” NotMP3downloader’s operator explains.

Then there’s the act of “stream-ripping” itself. While YouTube-MP3 downloaded digital content from YouTube using its own software, NotMP3downloader claims to do things differently.

“Our software doesn’t download any streaming content directly, but only launches a web browser with the video specified by a user. The capturing happens from a local machine’s sound card and doesn’t deal with any content streamed through a network,” its operator notes.

This part also seems quite important. YouTube-MP3 was accused of unlawfully circumventing technological measures implemented by YouTube to prevent people downloading or copying content. By opening up YouTube’s own website and viewing content in the way the site demands, NotMP3downloader says it does not “violate the website’s integrity nor performs direct download of audio or video files.”

Like the Betamax video recorder before it that enabled recording from analog TV, NotMP3downloader enables a user to record a YouTube stream on their local machine. This, its makers claim, means the software is completely legal and defeats all the claims made by the labels in the YouTube-MP3 lawsuit.

“What YouTube does is broadcasting content through the Internet. Thus, there is nothing wrong if users are allowed to watch such content later as they may want,” the NotMP3downloader team explain.

“It is worth noting that in Sony Corp. of America v. United City Studios, Inc. (464 U.S. 417) the United States Supreme Court held that such practice, also known as time-shifting, was lawful representing fair use under the US Copyright Act and causing no substantial harm to the copyright holder.”

While software that can record video and sounds locally are nothing new, the developments in the YouTube-MP3 case and this response from NotMP3downloader raises interesting questions.

We put some of them to none other than former RIAA Executive Vice President, Neil Turkewitz, who now works as President of Turkewitz Consulting Group.

Turkewitz stressed that he doesn’t speak for the industry as a whole or indeed the RIAA but it’s clear that his passion for protecting creators persists. He told us that in this instance, reliance on the Betamax decision is “misplaced”.

“The content is different, the activity is different, and the function is different,” Turkewitz told TF.

“The Sony decision must be understood in its context — the time shifting of audiovisual programming being broadcast from point to multipoint. The making available of content by a point-to-point interactive service like YouTube isn’t broadcasting — or at a minimum, is not a form of broadcasting akin to that considered by the Supreme Court in Sony.

“More fundamentally, broadcasting (right of communication to the public) is one of only several rights implicated by the service. And of course, issues of liability will be informed by considerations of purpose, effect and perceived harm. A court’s judgment will also be affected by whether it views the ‘innovation’ as an attempt to circumvent the requirements of law. The decision of the Supreme Court in ABC v. Aereo is certainly instructive in that regard.”

And there are other issues too. While YouTube itself is yet to take any legal action to deter users from downloading rather than merely streaming content, its terms of service are quite specific and seem to cover all eventualities.

“[Y]ou agree not to access Content or any reason other than your personal, non-commercial use solely as intended through and permitted by the normal functionality of the Service, and solely for Streaming,” YouTube’s ToS reads.

“‘Streaming’ means a contemporaneous digital transmission of the material by YouTube via the Internet to a user operated Internet enabled device in such a manner that the data is intended for real-time viewing and not intended to be downloaded (either permanently or temporarily), copied, stored, or redistributed by the user.

“You shall not copy, reproduce, distribute, transmit, broadcast, display, sell, license, or otherwise exploit any Content for any other purposes without the prior written consent of YouTube or the respective licensors of the Content.”

In this respect, it seems that a user doing anything but real-time streaming of YouTube content is breaching YouTube’s terms of service. The big question then, of course, is whether providing a tool specifically for that purpose represents an infringement of copyright.

The people behind Free MP3 Recorder believe that the “scope of application depends entirely on the end users’ intentions” which seems like a fair argument at first view. But, as usual, copyright law is incredibly complex and there are plenty of opposing views.

We asked the BPI, which took action against YouTubeMP3, for its take on this type of tool. The official response was “No comment” which doesn’t really clarify the position, at least for now.

Needless to say, the Betamax decision – relevant or not – doesn’t apply in the UK. But that only adds more parameters into the mix – and perhaps more opportunities for lawyers to make money arguing for and against tools like this in the future.

Source: TF, for the latest info on copyright, file-sharing, torrent sites and more. We also have VPN reviews, discounts, offers and coupons.

Tracking Cookies and GDPR

Post Syndicated from Bozho original https://techblog.bozho.net/tracking-cookies-gdpr/

GDPR is the new data protection regulation, as you probably already know. I’ve given a detailed practical advice for what it means for developers (and product owners). However, there’s one thing missing there – cookies. The elephant in the room.

Previously I’ve stated that cookies are subject to another piece of legislation – the ePrivacy directive, which is getting updated and its new version will be in force a few years from now. And while that’s technically correct, cookies seem to be affected by GDPR as well. In a way I’ve underestimated that effect.

When you do a Google search on “GDPR cookies”, you’ll pretty quickly realize that a) there’s not too much information and b) there’s not much technical understanding of the issue.

What appears to be the consensus is that GDPR does change the way cookies are handled. More specifically – tracking cookies. Here’s recital 30:

(30) Natural persons may be associated with online identifiers provided by their devices, applications, tools and protocols, such as internet protocol addresses, cookie identifiers or other identifiers such as radio frequency identification tags. This may leave traces which, in particular when combined with unique identifiers and other information received by the servers, may be used to create profiles of the natural persons and identify them.

How tracking cookies work – a 3rd party (usually an ad network) gives you a code snippet that you place on your website, for example to display ads. That code snippet, however, calls “home” (makes a request to the 3rd party domain). If the 3rd party has previously been used on your computer, it has created a cookie. In the example of Facebook, they have the cookie with your Facebook identifier because you’ve logged in to Facebook. So this cookie (with your identifier) is sent with the request. The request also contains all the details from the page. In effect, you are uniquely identified by an identifier (in the case of Facebook and Google – fully identified, rather than some random anonymous identifier as with other ad networks).

Your behaviour on the website is personal data. It gets associated with your identifier, which in turn is associated with your profile. And all of that is personal data. Who is responsible for collecting the website behaviour data, i.e. who is the “controller”? Is it Facebook (or any other 3rd party) that technically does the collection? No, it’s the website owner, as the behaviour data is obtained on their website, and they have put the tracking piece of code there. So they bear responsibility.

What’s the responsibility? So far it boiled down to displaying the useless “we use cookies” warning that nobody cares about. And the current (old) ePrivacy directive and its interpretations says that this is enough – if the users actions can unambiguously mean that they are fine with cookies – i.e. if they continue to use the website after seeing the warning – then you’re fine. This is no longer true from a GDPR perspective – you are collecting user data and you have to have a lawful ground for processing.

For the data collected by tracking cookies you have two options – “consent” and “legitimate interest”. Legitimate interest will be hard to prove – it is not something that a user reasonably expects, it is not necessary for you to provide the service. If your lawyers can get that option to fly, good for them, but I’m not convinced regulators will be happy with that.

The other option is “consent”. You have to ask your users explicitly – that means “with a checkbox” – to let you use tracking cookies. That has two serious implications – from technical and usability point of view.

  • The technical issue is that the data is sent via 3rd party code as soon as the page loads and before the user can give their consent. And that’s already a violation. You can, of course, have the 3rd party code be dynamically inserted only after the user gives consent, but that will require some fiddling with javascript and might not always work depending on the provider. And you’d have to support opt-out at any time (which would in turn disable the 3rd party snippet). It would require actual coding, rather than just copy-pasting a snippet.
  • The usability aspect is the bigger issue – while you could neatly tuck a cookie warning at the bottom, you’d now have to have a serious, “stop the world” popup that asks for consent if you want anyone to click it. You can, of course, just add a checkbox to the existing cookie warning, but don’t expect anyone to click it.

These aspects pose a significant questions: is it worth it to have tracking cookies? Is developing new functionality worth it, is interrupting the user worth it, and is implementing new functionality just so that users never clicks a hidden checkbox worth it? Especially given that Firefox now blocks all tracking cookies and possibly other browsers will follow?

That by itself is an interesting topic – Firefox has basically implemented the most strict form of requirements of the upcoming ePrivacy directive update (that would turn it into an ePrivacy regulation). Other browsers will have to follow, even though Google may not be happy to block their own tracking cookies. I hope other browsers follow Firefox in tracking protection and the issue will be gone automatically.

To me it seems that it will be increasingly not worthy to have tracking cookies on your website. They add regulatory obligations for you and give you very little benefit (yes, you could track engagement from ads, but you can do that in other ways, arguably by less additional code than supporting the cookie consents). And yes, the cookie consent will be “outsourced” to browsers after the ePrivacy regulation is passed, but we can’t be sure at the moment whether there won’t be technical whack-a-mole between browsers and advertisers and whether you wouldn’t still need additional effort to have dynamic consent for tracking cookies. (For example there are reported issues that Firefox used to make Facebook login fail if tracking protection is enabled. Which could be a simple bug, or could become a strategy by big vendors in the future to force browsers into a less strict tracking protection).

Okay, we’ve decided it’s not worth it managing tracking cookies. But do you have a choice as a website owner? Can you stop your ad network from using them? (Remember – you are liable if users’ data is collected by visiting your website). And currently the answer is no – you can’t disable that. You can’t have “just the ads”. This is part of the “deal” – you get money for the ads you place, but you participate in a big “surveillance” network. Users have a way to opt out (e.g. Google AdWords gives them that option). You, as a website owner, don’t.

Facebook has a recommendations page that says “you take care of getting the consent”. But for example the “like button” plugin doesn’t have an option to not send any data to Facebook.

And sometimes you don’t want to serve ads, just track user behaviour and measure conversion. But even if you ask for consent for that and conditionally insert the plugin/snippet, do you actually know what data it sends? And what it’s used for? Because you have to know in order to inform your users. “Do you agree to use tracking cookies that Facebook has inserted in order to collect data about your behaviour on our website” doesn’t sound compelling.

So, what to do? The easiest thing is just not to use any 3rd party ad-related plugins. But that’s obviously not an option, as ad revenue is important, especially in the publishing industry. I don’t have a good answer, apart from “Regulators should pressure ad networks to provide opt-outs and clearly document their data usage”. They have to do that under GDPR, and while website owners are responsible for their users’ data, the ad networks that are in the role of processors in this case (as you delegate the data collection for your visitors to them) also have obligation to assist you in fulfilling your obligations. So ask Facebook – what should I do with your tracking cookies? And when the regulator comes after a privacy-aware customer files a complaint, you could prove that you’ve tried.

The ethical debate whether it’s wrong to collect data about peoples’ behaviour without their informed consent is an easy one. And that’s why I don’t put blame on the regulators – they are putting the ethical consensus in law. It gets more complicated if not allowing tracking means some internet services are no longer profitable and therefore can’t exist. Can we have the cake and eat it too?

The post Tracking Cookies and GDPR appeared first on Bozho's tech blog.

Reactive Microservices Architecture on AWS

Post Syndicated from Sascha Moellering original https://aws.amazon.com/blogs/architecture/reactive-microservices-architecture-on-aws/

Microservice-application requirements have changed dramatically in recent years. These days, applications operate with petabytes of data, need almost 100% uptime, and end users expect sub-second response times. Typical N-tier applications can’t deliver on these requirements.

Reactive Manifesto, published in 2014, describes the essential characteristics of reactive systems including: responsiveness, resiliency, elasticity, and being message driven.

Being message driven is perhaps the most important characteristic of reactive systems. Asynchronous messaging helps in the design of loosely coupled systems, which is a key factor for scalability. In order to build a highly decoupled system, it is important to isolate services from each other. As already described, isolation is an important aspect of the microservices pattern. Indeed, reactive systems and microservices are a natural fit.

Implemented Use Case
This reference architecture illustrates a typical ad-tracking implementation.

Many ad-tracking companies collect massive amounts of data in near-real-time. In many cases, these workloads are very spiky and heavily depend on the success of the ad-tech companies’ customers. Typically, an ad-tracking-data use case can be separated into a real-time part and a non-real-time part. In the real-time part, it is important to collect data as fast as possible and ask several questions including:,  “Is this a valid combination of parameters?,””Does this program exist?,” “Is this program still valid?”

Because response time has a huge impact on conversion rate in advertising, it is important for advertisers to respond as fast as possible. This information should be kept in memory to reduce communication overhead with the caching infrastructure. The tracking application itself should be as lightweight and scalable as possible. For example, the application shouldn’t have any shared mutable state and it should use reactive paradigms. In our implementation, one main application is responsible for this real-time part. It collects and validates data, responds to the client as fast as possible, and asynchronously sends events to backend systems.

The non-real-time part of the application consumes the generated events and persists them in a NoSQL database. In a typical tracking implementation, clicks, cookie information, and transactions are matched asynchronously and persisted in a data store. The matching part is not implemented in this reference architecture. Many ad-tech architectures use frameworks like Hadoop for the matching implementation.

The system can be logically divided into the data collection partand the core data updatepart. The data collection part is responsible for collecting, validating, and persisting the data. In the core data update part, the data that is used for validation gets updated and all subscribers are notified of new data.

Components and Services

Main Application
The main application is implemented using Java 8 and uses Vert.x as the main framework. Vert.x is an event-driven, reactive, non-blocking, polyglot framework to implement microservices. It runs on the Java virtual machine (JVM) by using the low-level IO library Netty. You can write applications in Java, JavaScript, Groovy, Ruby, Kotlin, Scala, and Ceylon. The framework offers a simple and scalable actor-like concurrency model. Vert.x calls handlers by using a thread known as an event loop. To use this model, you have to write code known as “verticles.” Verticles share certain similarities with actors in the actor model. To use them, you have to implement the verticle interface. Verticles communicate with each other by generating messages in  a single event bus. Those messages are sent on the event bus to a specific address, and verticles can register to this address by using handlers.

With only a few exceptions, none of the APIs in Vert.x block the calling thread. Similar to Node.js, Vert.x uses the reactor pattern. However, in contrast to Node.js, Vert.x uses several event loops. Unfortunately, not all APIs in the Java ecosystem are written asynchronously, for example, the JDBC API. Vert.x offers a possibility to run this, blocking APIs without blocking the event loop. These special verticles are called worker verticles. You don’t execute worker verticles by using the standard Vert.x event loops, but by using a dedicated thread from a worker pool. This way, the worker verticles don’t block the event loop.

Our application consists of five different verticles covering different aspects of the business logic. The main entry point for our application is the HttpVerticle, which exposes an HTTP-endpoint to consume HTTP-requests and for proper health checking. Data from HTTP requests such as parameters and user-agent information are collected and transformed into a JSON message. In order to validate the input data (to ensure that the program exists and is still valid), the message is sent to the CacheVerticle.

This verticle implements an LRU-cache with a TTL of 10 minutes and a capacity of 100,000 entries. Instead of adding additional functionality to a standard JDK map implementation, we use Google Guava, which has all the features we need. If the data is not in the L1 cache, the message is sent to the RedisVerticle. This verticle is responsible for data residing in Amazon ElastiCache and uses the Vert.x-redis-client to read data from Redis. In our example, Redis is the central data store. However, in a typical production implementation, Redis would just be the L2 cache with a central data store like Amazon DynamoDB. One of the most important paradigms of a reactive system is to switch from a pull- to a push-based model. To achieve this and reduce network overhead, we’ll use Redis pub/sub to push core data changes to our main application.

Vert.x also supports direct Redis pub/sub-integration, the following code shows our subscriber-implementation:

vertx.eventBus().<JsonObject>consumer(REDIS_PUBSUB_CHANNEL_VERTX, received -> {

JsonObject value = received.body().getJsonObject("value");

String message = value.getString("message");

JsonObject jsonObject = new JsonObject(message);

eb.send(CACHE_REDIS_EVENTBUS_ADDRESS, jsonObject);

});

redis.subscribe(Constants.REDIS_PUBSUB_CHANNEL, res -> {

if (res.succeeded()) {

LOGGER.info("Subscribed to " + Constants.REDIS_PUBSUB_CHANNEL);

} else {

LOGGER.info(res.cause());

}

});

The verticle subscribes to the appropriate Redis pub/sub-channel. If a message is sent over this channel, the payload is extracted and forwarded to the cache-verticle that stores the data in the L1-cache. After storing and enriching data, a response is sent back to the HttpVerticle, which responds to the HTTP request that initially hit this verticle. In addition, the message is converted to ByteBuffer, wrapped in protocol buffers, and send to an Amazon Kinesis Data Stream.

The following example shows a stripped-down version of the KinesisVerticle:

public class KinesisVerticle extends AbstractVerticle {

private static final Logger LOGGER = LoggerFactory.getLogger(KinesisVerticle.class);

private AmazonKinesisAsync kinesisAsyncClient;

private String eventStream = "EventStream";

@Override

public void start() throws Exception {

EventBus eb = vertx.eventBus();

kinesisAsyncClient = createClient();

eventStream = System.getenv(STREAM_NAME) == null ? "EventStream" : System.getenv(STREAM_NAME);

eb.consumer(Constants.KINESIS_EVENTBUS_ADDRESS, message -> {

try {

TrackingMessage trackingMessage = Json.decodeValue((String)message.body(), TrackingMessage.class);

String partitionKey = trackingMessage.getMessageId();

byte [] byteMessage = createMessage(trackingMessage);

ByteBuffer buf = ByteBuffer.wrap(byteMessage);

sendMessageToKinesis(buf, partitionKey);

message.reply("OK");

}

catch (KinesisException exc) {

LOGGER.error(exc);

}

});

}

Kinesis Consumer
This AWS Lambda function consumes data from an Amazon Kinesis Data Stream and persists the data in an Amazon DynamoDB table. In order to improve testability, the invocation code is separated from the business logic. The invocation code is implemented in the class KinesisConsumerHandler and iterates over the Kinesis events pulled from the Kinesis stream by AWS Lambda. Each Kinesis event is unwrapped and transformed from ByteBuffer to protocol buffers and converted into a Java object. Those Java objects are passed to the business logic, which persists the data in a DynamoDB table. In order to improve duration of successive Lambda calls, the DynamoDB-client is instantiated lazily and reused if possible.

Redis Updater
From time to time, it is necessary to update core data in Redis. A very efficient implementation for this requirement is using AWS Lambda and Amazon Kinesis. New core data is sent over the AWS Kinesis stream using JSON as data format and consumed by a Lambda function. This function iterates over the Kinesis events pulled from the Kinesis stream by AWS Lambda. Each Kinesis event is unwrapped and transformed from ByteBuffer to String and converted into a Java object. The Java object is passed to the business logic and stored in Redis. In addition, the new core data is also sent to the main application using Redis pub/sub in order to reduce network overhead and converting from a pull- to a push-based model.

The following example shows the source code to store data in Redis and notify all subscribers:

public void updateRedisData(final TrackingMessage trackingMessage, final Jedis jedis, final LambdaLogger logger) {

try {

ObjectMapper mapper = new ObjectMapper();

String jsonString = mapper.writeValueAsString(trackingMessage);

Map<String, String> map = marshal(jsonString);

String statusCode = jedis.hmset(trackingMessage.getProgramId(), map);

}

catch (Exception exc) {

if (null == logger)

exc.printStackTrace();

else

logger.log(exc.getMessage());

}

}

public void notifySubscribers(final TrackingMessage trackingMessage, final Jedis jedis, final LambdaLogger logger) {

try {

ObjectMapper mapper = new ObjectMapper();

String jsonString = mapper.writeValueAsString(trackingMessage);

jedis.publish(Constants.REDIS_PUBSUB_CHANNEL, jsonString);

}

catch (final IOException e) {

log(e.getMessage(), logger);

}

}

Similarly to our Kinesis Consumer, the Redis-client is instantiated somewhat lazily.

Infrastructure as Code
As already outlined, latency and response time are a very critical part of any ad-tracking solution because response time has a huge impact on conversion rate. In order to reduce latency for customers world-wide, it is common practice to roll out the infrastructure in different AWS Regions in the world to be as close to the end customer as possible. AWS CloudFormation can help you model and set up your AWS resources so that you can spend less time managing those resources and more time focusing on your applications that run in AWS.

You create a template that describes all the AWS resources that you want (for example, Amazon EC2 instances or Amazon RDS DB instances), and AWS CloudFormation takes care of provisioning and configuring those resources for you. Our reference architecture can be rolled out in different Regions using an AWS CloudFormation template, which sets up the complete infrastructure (for example, Amazon Virtual Private Cloud (Amazon VPC), Amazon Elastic Container Service (Amazon ECS) cluster, Lambda functions, DynamoDB table, Amazon ElastiCache cluster, etc.).

Conclusion
In this blog post we described reactive principles and an example architecture with a common use case. We leveraged the capabilities of different frameworks in combination with several AWS services in order to implement reactive principles—not only at the application-level but also at the system-level. I hope I’ve given you ideas for creating your own reactive applications and systems on AWS.

About the Author

Sascha Moellering is a Senior Solution Architect. Sascha is primarily interested in automation, infrastructure as code, distributed computing, containers and JVM. He can be reached at [email protected]

 

 

Simplify Querying Nested JSON with the AWS Glue Relationalize Transform

Post Syndicated from Trevor Roberts original https://aws.amazon.com/blogs/big-data/simplify-querying-nested-json-with-the-aws-glue-relationalize-transform/

AWS Glue has a transform called Relationalize that simplifies the extract, transform, load (ETL) process by converting nested JSON into columns that you can easily import into relational databases. Relationalize transforms the nested JSON into key-value pairs at the outermost level of the JSON document. The transformed data maintains a list of the original keys from the nested JSON separated by periods.

Let’s look at how Relationalize can help you with a sample use case.

An example of Relationalize in action

Suppose that the developers of a video game want to use a data warehouse like Amazon Redshift to run reports on player behavior based on data that is stored in JSON. Sample 1 shows example user data from the game. The player named “user1” has characteristics such as race, class, and location in nested JSON data. Further down, the player’s arsenal information includes additional nested JSON data. If the developers want to ETL this data into their data warehouse, they might have to resort to nested loops or recursive functions in their code.

Sample 1: Nested JSON

{
	"player": {
		"username": "user1",
		"characteristics": {
			"race": "Human",
			"class": "Warlock",
			"subclass": "Dawnblade",
			"power": 300,
			"playercountry": "USA"
		},
		"arsenal": {
			"kinetic": {
				"name": "Sweet Business",
				"type": "Auto Rifle",
				"power": 300,
				"element": "Kinetic"
			},
			"energy": {
				"name": "MIDA Mini-Tool",
				"type": "Submachine Gun",
				"power": 300,
				"element": "Solar"
			},
			"power": {
				"name": "Play of the Game",
				"type": "Grenade Launcher",
				"power": 300,
				"element": "Arc"
			}
		},
		"armor": {
			"head": "Eye of Another World",
			"arms": "Philomath Gloves",
			"chest": "Philomath Robes",
			"leg": "Philomath Boots",
			"classitem": "Philomath Bond"
		},
		"location": {
			"map": "Titan",
			"waypoint": "The Rig"
		}
	}
}

Instead, the developers can use the Relationalize transform. Sample 2 shows what the transformed data looks like.

Sample 2: Flattened JSON

{
    "player.username": "user1",
    "player.characteristics.race": "Human",
    "player.characteristics.class": "Warlock",
    "player.characteristics.subclass": "Dawnblade",
    "player.characteristics.power": 300,
    "player.characteristics.playercountry": "USA",
    "player.arsenal.kinetic.name": "Sweet Business",
    "player.arsenal.kinetic.type": "Auto Rifle",
    "player.arsenal.kinetic.power": 300,
    "player.arsenal.kinetic.element": "Kinetic",
    "player.arsenal.energy.name": "MIDA Mini-Tool",
    "player.arsenal.energy.type": "Submachine Gun",
    "player.arsenal.energy.power": 300,
    "player.arsenal.energy.element": "Solar",
    "player.arsenal.power.name": "Play of the Game",
    "player.arsenal.power.type": "Grenade Launcher",
    "player.arsenal.power.power": 300,
    "player.arsenal.power.element": "Arc",
    "player.armor.head": "Eye of Another World",
    "player.armor.arms": "Philomath Gloves",
    "player.armor.chest": "Philomath Robes",
    "player.armor.leg": "Philomath Boots",
    "player.armor.classitem": "Philomath Bond",
    "player.location.map": "Titan",
    "player.location.waypoint": "The Rig"
}

You can then write the data to a database or to a data warehouse. You can also write it to delimited text files, such as in comma-separated value (CSV) format, or columnar file formats such as Optimized Row Columnar (ORC) format. You can use either of these format types for long-term storage in Amazon S3. Storing the transformed files in S3 provides the additional benefit of being able to query this data using Amazon Athena or Amazon Redshift Spectrum. You can further extend the usefulness of the data by performing joins between data stored in S3 and the data stored in an Amazon Redshift data warehouse.

Before we get started…

In my example, I took two preparatory steps that save some time in your ETL code development:

  1. I stored my data in an Amazon S3 bucket and used an AWS Glue crawler to make my data available in the AWS Glue data catalog. You can find instructions on how to do that in Cataloging Tables with a Crawler in the AWS Glue documentation. The AWS Glue database name I used was “blog,” and the table name was “players.” You can see these values in use in the sample code that follows.
  2. I deployed a Zeppelin notebook using the automated deployment available within AWS Glue. If you already used an AWS Glue development endpoint to deploy a Zeppelin notebook, you can skip the deployment instructions. Otherwise, let’s quickly review how to deploy Zeppelin.

Deploying a Zeppelin notebook with AWS Glue

The following steps are outlined in the AWS Glue documentation, and I include a few screenshots here for clarity.

First, create two IAM roles:

Next, in the AWS Glue Management Console, choose Dev endpoints, and then choose Add endpoint.

Specify a name for the endpoint and the AWS Glue IAM role that you created.

On the networking screen, choose Skip Networking because our code only communicates with S3.

Complete the development endpoint process by providing a Secure Shell (SSH) public key and confirming your settings.

When your new development endpoint’s Provisioning status changes from PROVISIONING to READY, choose your endpoint, and then for Actions choose Create notebook server.

Enter the notebook server details, including the role you previously created and a security group with inbound access allowed on TCP port 443.

Doing this automatically launches an AWS CloudFormation template. The output specifies the URL that you can use to access your Zeppelin notebook with the username and password you specified in the wizard.

How do we flatten nested JSON?

With my data loaded and my notebook server ready, I accessed Zeppelin, created a new note, and set my interpreter to spark. I used some Python code that AWS Glue previously generated for another job that outputs to ORC. Then I added the Relationalize transform. You can see the resulting Python code in Sample 3.­

Sample 3: Python code to transform the nested JSON and output it to ORC

import sys
from awsglue.transforms import *
from awsglue.utils import getResolvedOptions
from pyspark.context import SparkContext
from awsglue.context import GlueContext
from awsglue.job import Job
#from awsglue.transforms import Relationalize

# Begin variables to customize with your information
glue_source_database = "blog"
glue_source_table = "players"
glue_temp_storage = "s3://blog-example-edz/temp"
glue_relationalize_output_s3_path = "s3://blog-example-edz/output-flat"
dfc_root_table_name = "root" #default value is "roottable"
# End variables to customize with your information

glueContext = GlueContext(spark.sparkContext)
datasource0 = glueContext.create_dynamic_frame.from_catalog(database = glue_source_database, table_name = glue_source_table, transformation_ctx = "datasource0")
dfc = Relationalize.apply(frame = datasource0, staging_path = glue_temp_storage, name = dfc_root_table_name, transformation_ctx = "dfc")
blogdata = dfc.select(dfc_root_table_name)
blogdataoutput = glueContext.write_dynamic_frame.from_options(frame = blogdata, connection_type = "s3", connection_options = {"path": glue_relationalize_output_s3_path}, format = "orc", transformation_ctx = "blogdataoutput")

What exactly is going on in this script?

After the import statements, we instantiate a GlueContext object, which allows us to work with the data in AWS Glue. Next, we create a DynamicFrame (datasource0) from the “players” table in the AWS Glue “blog” database. We use this DynamicFrame to perform any necessary operations on the data structure before it’s written to our desired output format. The source files remain unchanged.

We then run the Relationalize transform (Relationalize.apply()) with our datasource0 as one of the parameters. Another important parameter is the name parameter, which is a key that identifies our data after the transformation completes.

The Relationalize.apply() method returns a DynamicFrameCollection, and this is stored in the dfc variable. Before we can write our data to S3, we need to select the DynamicFrame from the DynamicFrameCollection object. We do this with the dfc.select() method. The correct DynamicFrame is stored in the blogdata variable.

You might be curious why a DynamicFrameCollection was returned when we started with a single DynamicFrame. This return value comes from the way Relationalize treats arrays in the JSON document: A DynamicFrame is created for each array. Together with the root data structure, each generated DynamicFrame is added to a DynamicFrameCollection when Relationalize completes its work. Although we didn’t have any arrays in our data, it’s good to keep this in mind. Finally, we output (blogdataoutput) the root DynamicFrame to ORC files in S3.

Using the transformed data

One of the use cases we discussed earlier was using Amazon Athena or Amazon Redshift Spectrum to query the ORC files.

I used the following SQL DDL statements to create external tables in both services to enable queries of my data stored in Amazon S3.

Sample 4: Amazon Athena DDL

CREATE EXTERNAL TABLE IF NOT EXISTS blog.blog_data_athena_test (
  `characteristics_race` string,
  `characteristics_class` string,
  `characteristics_subclass` string,
  `characteristics_power` int,
  `characteristics_playercountry` string,
  `kinetic_name` string,
  `kinetic_type` string,
  `kinetic_power` int,
  `kinetic_element` string,
  `energy_name` string,
  `energy_type` string,
  `energy_power` int,
  `energy_element` string,
  `power_name` string,
  `power_type` string,
  `power_power` int,
  `power_element` string,
  `armor_head` string,
  `armor_arms` string,
  `armor_chest` string,
  `armor_leg` string,
  `armor_classitem` string,
  `map` string,
  `waypoint` string 
)
ROW FORMAT SERDE 'org.apache.hadoop.hive.ql.io.orc.OrcSerde'
WITH SERDEPROPERTIES (
  'serialization.format' = '1'
) LOCATION 's3://blog-example-edz/output-flat/'
TBLPROPERTIES ('has_encrypted_data'='false');

 

Sample 5: Amazon Redshift Spectrum DDL

-- Create a Schema
-- A single schema can be used with multiple external tables.
-- This step is only required once for the external tables you create.
create external schema spectrum 
from data catalog 
database 'blog' 
iam_role 'arn:aws:iam::0123456789:role/redshift-role'
create external database if not exists;

-- Create an external table in the schema
create external table spectrum.blog(
  username VARCHAR,
  characteristics_race VARCHAR,
  characteristics_class VARCHAR,
  characteristics_subclass VARCHAR,
  characteristics_power INTEGER,
  characteristics_playercountry VARCHAR,
  kinetic_name VARCHAR,
  kinetic_type VARCHAR,
  kinetic_power INTEGER,
  kinetic_element VARCHAR,
  energy_name VARCHAR,
  energy_type VARCHAR,
  energy_power INTEGER,
  energy_element VARCHAR,
  power_name VARCHAR,
  power_type VARCHAR,
  power_power INTEGER,
  power_element VARCHAR,
  armor_head VARCHAR,
  armor_arms VARCHAR,
  armor_chest VARCHAR,
  armor_leg VARCHAR,
  armor_classItem VARCHAR,
  map VARCHAR,
  waypoint VARCHAR)
stored as orc
location 's3://blog-example-edz/output-flat';

I even ran a query, shown in Sample 6, that joined my Redshift Spectrum table (spectrum.playerdata) with data in an Amazon Redshift table (public.raids) to generate advanced reports. In the where clause, I join the two tables based on the username values that are common to both data sources.

Sample 6: Select statement with a join of Redshift Spectrum data with Amazon Redshift data

-- Get Total Raid Completions for the Hunter Class.
select spectrum.playerdata.characteristics_class as class, sum(public.raids."completions.val.raids.leviathan") as "Total Hunter Leviathan Raid Completions" from spectrum.playerdata, public.raids
where spectrum.playerdata.username = public.raids."completions.val.username"
and spectrum.playerdata.characteristics_class = 'Hunter'
group by spectrum.playerdata.characteristics_class;

Summary

This post demonstrated how simple it can be to flatten nested JSON data with AWS Glue, using the Relationalize transform to automate the conversion of nested JSON. AWS Glue also automates the deployment of Zeppelin notebooks that you can use to develop your Python automation script. Finally, AWS Glue can output the transformed data directly to a relational database, or to files in Amazon S3 for further analysis with tools such as Amazon Athena and Amazon Redshift Spectrum.

As great as Relationalize is, it’s not the only transform available with AWS Glue. You can see a complete list of available transforms in Built-In Transforms in the AWS Glue documentation. Try them out today!


Additional Reading

If you found this post useful, be sure to check out Using Amazon Redshift Spectrum, Amazon Athena and AWS Glue with Node.js in Production and Build a Data Lake Foundation with AWS Glue and Amazon S3.


About the Author

Trevor Roberts Jr is a Solutions Architect with AWS. He provides architectural guidance to help customers achieve success in the cloud. In his spare time, Trevor enjoys traveling to new places and spending time with family.

Marvellous retrofitted home assistants

Post Syndicated from Alex Bate original https://www.raspberrypi.org/blog/retrofitted-home-assistants/

As more and more digital home assistants are appearing on the consumer market, it’s not uncommon to see the towering Amazon Echo or sleek Google Home when visiting friends or family. But we, the maker community, are rarely happy unless our tech stands out from the rest. So without further ado, here’s a roundup of some fantastic retrofitted home assistant projects you can recreate and give pride of place in your kitchen, on your bookshelf, or wherever else you’d like to talk to your virtual, disembodied PA.

Google AIY Robot Conversion

Turned an 80s Tomy Mr Money into a little Google AIY / Raspberry Pi based assistant.

Matt ‘Circuitbeard’ Brailsford’s Tomy Mr Money Google AIY Assistant is just one of many home-brew home assistants makers have built since the release of APIs for Amazon Alexa and Google Home. Here are some more…

Teddy Ruxpin

Oh Teddy, how exciting and mysterious you were when I unwrapped you back in the mideighties. With your awkwardly moving lips and twitching eyelids, you were the cream of the crop of robotic toys! How was I to know that during my thirties, you would become augmented with home assistant software and suddenly instil within me a fear unlike any I’d felt before? (Save for my lifelong horror of ET…)

Alexa Ruxpin – Raspberry Pi & Alexa Powered Teddy Bear

Please watch: “DIY Fidget LED Display – Part 1” https://www.youtube.com/watch?v=FAZIc82Duzk -~-~~-~~~-~~-~- There are tons of virtual assistants out on the market: Siri, Ok Google, Alexa, etc. I had this crazy idea…what if I made the virtual assistant real…kinda. I decided to take an old animatronic teddy bear and hack it so that it ran Amazon Alexa.

Several makers around the world have performed surgery on Teddy to install a Raspberry Pi within his stomach and integrate him with Amazon Alexa Voice or Google’s AIY Projects Voice kit. And because these makers are talented, they’ve also managed to hijack Teddy’s wiring to make his lips move in time with his responses to your commands. Freaky…

Speaking of freaky: check out Zack’s Furlexa — an Amazon Alexa Furby that will haunt your nightmares.

Give old tech new life

Devices that were the height of technology when you purchased them may now be languishing in your attic collecting dust. With new and improved versions of gadgets and gizmos being released almost constantly, it is likely that your household harbours a spare whosit or whatsit which you can dismantle and give a new Raspberry Pi heart and purpose.

Take, for example, Martin Mander’s Google Pi intercom. By gutting and thoroughly cleaning a vintage intercom, Martin fashioned a suitable housing the Google AIY Projects Voice kit to create a new home assistant for his house:

1986 Google Pi Intercom

This is a 1986 Radio Shack Intercom that I’ve converted into a Google Home style device using a Raspberry Pi and the Google AIY (Artificial Intelligence Yourself) kit that came free with the MagPi magazine (issue 57). It uses the Google Assistant to answer questions and perform actions, using IFTTT to integrate with smart home accessories and other web services.

Not only does this build look fantastic, it’s also a great conversation starter for any visitors who had a similar device during the eighties.

Also take a look at Martin’s 1970s Amazon Alexa phone for more nostalgic splendour.

Put it in a box

…and then I’ll put that box inside of another box, and then I’ll mail that box to myself, and when it arrives…

A GIF from the emperors new groove - Raspberry Pi Home Assistant

A GIF. A harmless, little GIF…and proof of the comms team’s obsession with The Emperor’s New Groove.

You don’t have to be fancy when it comes to housing your home assistant. And often, especially if you’re working with the smaller people in your household, the results of a simple homespun approach are just as delightful.

Here are Hannah and her dad Tom, explaining how they built a home assistant together and fit it inside an old cigar box:

Raspberry Pi 3 Amazon Echo – The Alexa Kids Build!

My 7 year old daughter and I decided to play around with the Raspberry Pi and build ourselves an Amazon Echo (Alexa). The video tells you about what we did and the links below will take you to all the sites we used to get this up and running.

Also see the Google AIY Projects Voice kit — the cardboard box-est of home assistant boxes.

Make your own home assistant

And now it’s your turn! I challenge you all (and also myself) to create a home assistant using the Raspberry Pi. Whether you decide to fit Amazon Alexa inside an old shoebox or Google Home inside your sister’s Barbie, I’d love to see what you create using the free home assistant software available online.

Check out these other home assistants for Raspberry Pi, and keep an eye on our blog to see what I manage to create as part of the challenge.

Ten virtual house points for everyone who shares their build with us online, either in the comments below or by tagging us on your social media account.

The post Marvellous retrofitted home assistants appeared first on Raspberry Pi.

Object models

Post Syndicated from Eevee original https://eev.ee/blog/2017/11/28/object-models/

Anonymous asks, with dollars:

More about programming languages!

Well then!

I’ve written before about what I think objects are: state and behavior, which in practice mostly means method calls.

I suspect that the popular impression of what objects are, and also how they should work, comes from whatever C++ and Java happen to do. From that point of view, the whole post above is probably nonsense. If the baseline notion of “object” is a rigid definition woven tightly into the design of two massively popular languages, then it doesn’t even make sense to talk about what “object” should mean — it does mean the features of those languages, and cannot possibly mean anything else.

I think that’s a shame! It piles a lot of baggage onto a fairly simple idea. Polymorphism, for example, has nothing to do with objects — it’s an escape hatch for static type systems. Inheritance isn’t the only way to reuse code between objects, but it’s the easiest and fastest one, so it’s what we get. Frankly, it’s much closer to a speed tradeoff than a fundamental part of the concept.

We could do with more experimentation around how objects work, but that’s impossible in the languages most commonly thought of as object-oriented.

Here, then, is a (very) brief run through the inner workings of objects in four very dynamic languages. I don’t think I really appreciated objects until I’d spent some time with Python, and I hope this can help someone else whet their own appetite.

Python 3

Of the four languages I’m going to touch on, Python will look the most familiar to the Java and C++ crowd. For starters, it actually has a class construct.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
class Vector:
    def __init__(self, x, y):
        self.x = x
        self.y = y

    def __neg__(self):
        return Vector(-self.x, -self.y)

    def __div__(self, denom):
        return Vector(self.x / denom, self.y / denom)

    @property
    def magnitude(self):
        return (self.x ** 2 + self.y ** 2) ** 0.5

    def normalized(self):
        return self / self.magnitude

The __init__ method is an initializer, which is like a constructor but named differently (because the object already exists in a usable form by the time the initializer is called). Operator overloading is done by implementing methods with other special __dunder__ names. Properties can be created with @property, where the @ is syntax for applying a wrapper function to a function as it’s defined. You can do inheritance, even multiply:

1
2
3
4
class Foo(A, B, C):
    def bar(self, x, y, z):
        # do some stuff
        super().bar(x, y, z)

Cool, a very traditional object model.

Except… for some details.

Some details

For one, Python objects don’t have a fixed layout. Code both inside and outside the class can add or remove whatever attributes they want from whatever object they want. The underlying storage is just a dict, Python’s mapping type. (Or, rather, something like one. Also, it’s possible to change, which will probably be the case for everything I say here.)

If you create some attributes at the class level, you’ll start to get a peek behind the curtains:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
class Foo:
    values = []

    def add_value(self, value):
        self.values.append(value)

a = Foo()
b = Foo()
a.add_value('a')
print(a.values)  # ['a']
b.add_value('b')
print(b.values)  # ['a', 'b']

The [] assigned to values isn’t a default assigned to each object. In fact, the individual objects don’t know about it at all! You can use vars(a) to get at the underlying storage dict, and you won’t see a values entry in there anywhere.

Instead, values lives on the class, which is a value (and thus an object) in its own right. When Python is asked for self.values, it checks to see if self has a values attribute; in this case, it doesn’t, so Python keeps going and asks the class for one.

Python’s object model is secretly prototypical — a class acts as a prototype, as a shared set of fallback values, for its objects.

In fact, this is also how method calls work! They aren’t syntactically special at all, which you can see by separating the attribute lookup from the call.

1
2
3
print("abc".startswith("a"))  # True
meth = "abc".startswith
print(meth("a"))  # True

Reading obj.method looks for a method attribute; if there isn’t one on obj, Python checks the class. Here, it finds one: it’s a function from the class body.

Ah, but wait! In the code I just showed, meth seems to “know” the object it came from, so it can’t just be a plain function. If you inspect the resulting value, it claims to be a “bound method” or “built-in method” rather than a function, too. Something funny is going on here, and that funny something is the descriptor protocol.

Descriptors

Python allows attributes to implement their own custom behavior when read from or written to. Such an attribute is called a descriptor. I’ve written about them before, but here’s a quick overview.

If Python looks up an attribute, finds it in a class, and the value it gets has a __get__ method… then instead of using that value, Python will use the return value of its __get__ method.

The @property decorator works this way. The magnitude property in my original example was shorthand for doing this:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
class MagnitudeDescriptor:
    def __get__(self, instance, owner):
        if instance is None:
            return self
        return (instance.x ** 2 + instance.y ** 2) ** 0.5

class Vector:
    def __init__(self, x, y):
        self.x = x
        self.y = y

    magnitude = MagnitudeDescriptor()

When you ask for somevec.magnitude, Python checks somevec but doesn’t find magnitude, so it consults the class instead. The class does have a magnitude, and it’s a value with a __get__ method, so Python calls that method and somevec.magnitude evaluates to its return value. (The instance is None check is because __get__ is called even if you get the descriptor directly from the class via Vector.magnitude. A descriptor intended to work on instances can’t do anything useful in that case, so the convention is to return the descriptor itself.)

You can also intercept attempts to write to or delete an attribute, and do absolutely whatever you want instead. But note that, similar to operating overloading in Python, the descriptor must be on a class; you can’t just slap one on an arbitrary object and have it work.

This brings me right around to how “bound methods” actually work. Functions are descriptors! The function type implements __get__, and when a function is retrieved from a class via an instance, that __get__ bundles the function and the instance together into a tiny bound method object. It’s essentially:

1
2
3
4
5
class FunctionType:
    def __get__(self, instance, owner):
        if instance is None:
            return self
        return functools.partial(self, instance)

The self passed as the first argument to methods is not special or magical in any way. It’s built out of a few simple pieces that are also readily accessible to Python code.

Note also that because obj.method() is just an attribute lookup and a call, Python doesn’t actually care whether method is a method on the class or just some callable thing on the object. You won’t get the auto-self behavior if it’s on the object, but otherwise there’s no difference.

More attribute access, and the interesting part

Descriptors are one of several ways to customize attribute access. Classes can implement __getattr__ to intervene when an attribute isn’t found on an object; __setattr__ and __delattr__ to intervene when any attribute is set or deleted; and __getattribute__ to implement unconditional attribute access. (That last one is a fantastic way to create accidental recursion, since any attribute access you do within __getattribute__ will of course call __getattribute__ again.)

Here’s what I really love about Python. It might seem like a magical special case that descriptors only work on classes, but it really isn’t. You could implement exactly the same behavior yourself, in pure Python, using only the things I’ve just told you about. Classes are themselves objects, remember, and they are instances of type, so the reason descriptors only work on classes is that type effectively does this:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
class type:
    def __getattribute__(self, name):
        value = super().__getattribute__(name)
        # like all op overloads, __get__ must be on the type, not the instance
        ty = type(value)
        if hasattr(ty, '__get__'):
            # it's a descriptor!  this is a class access so there is no instance
            return ty.__get__(value, None, self)
        else:
            return value

You can even trivially prove to yourself that this is what’s going on by skipping over types behavior:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
class Descriptor:
    def __get__(self, instance, owner):
        print('called!')

class Foo:
    bar = Descriptor()

Foo.bar  # called!
type.__getattribute__(Foo, 'bar')  # called!
object.__getattribute__(Foo, 'bar')  # ...

And that’s not all! The mysterious super function, used to exhaustively traverse superclass method calls even in the face of diamond inheritance, can also be expressed in pure Python using these primitives. You could write your own superclass calling convention and use it exactly the same way as super.

This is one of the things I really like about Python. Very little of it is truly magical; virtually everything about the object model exists in the types rather than the language, which means virtually everything can be customized in pure Python.

Class creation and metaclasses

A very brief word on all of this stuff, since I could talk forever about Python and I have three other languages to get to.

The class block itself is fairly interesting. It looks like this:

1
2
class Name(*bases, **kwargs):
    # code

I’ve said several times that classes are objects, and in fact the class block is one big pile of syntactic sugar for calling type(...) with some arguments to create a new type object.

The Python documentation has a remarkably detailed description of this process, but the gist is:

  • Python determines the type of the new class — the metaclass — by looking for a metaclass keyword argument. If there isn’t one, Python uses the “lowest” type among the provided base classes. (If you’re not doing anything special, that’ll just be type, since every class inherits from object and object is an instance of type.)

  • Python executes the class body. It gets its own local scope, and any assignments or method definitions go into that scope.

  • Python now calls type(name, bases, attrs, **kwargs). The name is whatever was right after class; the bases are position arguments; and attrs is the class body’s local scope. (This is how methods and other class attributes end up on the class.) The brand new type is then assigned to Name.

Of course, you can mess with most of this. You can implement __prepare__ on a metaclass, for example, to use a custom mapping as storage for the local scope — including any reads, which allows for some interesting shenanigans. The only part you can’t really implement in pure Python is the scoping bit, which has a couple extra rules that make sense for classes. (In particular, functions defined within a class block don’t close over the class body; that would be nonsense.)

Object creation

Finally, there’s what actually happens when you create an object — including a class, which remember is just an invocation of type(...).

Calling Foo(...) is implemented as, well, a call. Any type can implement calls with the __call__ special method, and you’ll find that type itself does so. It looks something like this:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
# oh, a fun wrinkle that's hard to express in pure python: type is a class, so
# it's an instance of itself
class type:
    def __call__(self, *args, **kwargs):
        # remember, here 'self' is a CLASS, an instance of type.
        # __new__ is a true constructor: object.__new__ allocates storage
        # for a new blank object
        instance = self.__new__(self, *args, **kwargs)
        # you can return whatever you want from __new__ (!), and __init__
        # is only called on it if it's of the right type
        if isinstance(instance, self):
            instance.__init__(*args, **kwargs)
        return instance

Again, you can trivially confirm this by asking any type for its __call__ method. Assuming that type doesn’t implement __call__ itself, you’ll get back a bound version of types implementation.

1
2
>>> list.__call__
<method-wrapper '__call__' of type object at 0x7fafb831a400>

You can thus implement __call__ in your own metaclass to completely change how subclasses are created — including skipping the creation altogether, if you like.

And… there’s a bunch of stuff I haven’t even touched on.

The Python philosophy

Python offers something that, on the surface, looks like a “traditional” class/object model. Under the hood, it acts more like a prototypical system, where failed attribute lookups simply defer to a superclass or metaclass.

The language also goes to almost superhuman lengths to expose all of its moving parts. Even the prototypical behavior is an implementation of __getattribute__ somewhere, which you are free to completely replace in your own types. Proxying and delegation are easy.

Also very nice is that these features “bundle” well, by which I mean a library author can do all manner of convoluted hijinks, and a consumer of that library doesn’t have to see any of it or understand how it works. You only need to inherit from a particular class (which has a metaclass), or use some descriptor as a decorator, or even learn any new syntax.

This meshes well with Python culture, which is pretty big on the principle of least surprise. These super-advanced features tend to be tightly confined to single simple features (like “makes a weak attribute“) or cordoned with DSLs (e.g., defining a form/struct/database table with a class body). In particular, I’ve never seen a metaclass in the wild implement its own __call__.

I have mixed feelings about that. It’s probably a good thing overall that the Python world shows such restraint, but I wonder if there are some very interesting possibilities we’re missing out on. I implemented a metaclass __call__ myself, just once, in an entity/component system that strove to minimize fuss when communicating between components. It never saw the light of day, but I enjoyed seeing some new things Python could do with the same relatively simple syntax. I wouldn’t mind seeing, say, an object model based on composition (with no inheritance) built atop Python’s primitives.

Lua

Lua doesn’t have an object model. Instead, it gives you a handful of very small primitives for building your own object model. This is pretty typical of Lua — it’s a very powerful language, but has been carefully constructed to be very small at the same time. I’ve never encountered anything else quite like it, and “but it starts indexing at 1!” really doesn’t do it justice.

The best way to demonstrate how objects work in Lua is to build some from scratch. We need two key features. The first is metatables, which bear a passing resemblance to Python’s metaclasses.

Tables and metatables

The table is Lua’s mapping type and its primary data structure. Keys can be any value other than nil. Lists are implemented as tables whose keys are consecutive integers starting from 1. Nothing terribly surprising. The dot operator is sugar for indexing with a string key.

1
2
3
4
5
local t = { a = 1, b = 2 }
print(t['a'])  -- 1
print(t.b)  -- 2
t.c = 3
print(t['c'])  -- 3

A metatable is a table that can be associated with another value (usually another table) to change its behavior. For example, operator overloading is implemented by assigning a function to a special key in a metatable.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
local t = { a = 1, b = 2 }
--print(t + 0)  -- error: attempt to perform arithmetic on a table value

local mt = {
    __add = function(left, right)
        return 12
    end,
}
setmetatable(t, mt)
print(t + 0)  -- 12

Now, the interesting part: one of the special keys is __index, which is consulted when the base table is indexed by a key it doesn’t contain. Here’s a table that claims every key maps to itself.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
local t = {}
local mt = {
    __index = function(table, key)
        return key
    end,
}
setmetatable(t, mt)
print(t.foo)  -- foo
print(t.bar)  -- bar
print(t[3])  -- 3

__index doesn’t have to be a function, either. It can be yet another table, in which case that table is simply indexed with the key. If the key still doesn’t exist and that table has a metatable with an __index, the process repeats.

With this, it’s easy to have several unrelated tables that act as a single table. Call the base table an object, fill the __index table with functions and call it a class, and you have half of an object system. You can even get prototypical inheritance by chaining __indexes together.

At this point things are a little confusing, since we have at least three tables going on, so here’s a diagram. Keep in mind that Lua doesn’t actually have anything called an “object”, “class”, or “method” — those are just convenient nicknames for a particular structure we might build with Lua’s primitives.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
                    ╔═══════════╗        ...
                    ║ metatable ║         ║
                    ╟───────────╢   ┌─────╨───────────────────────┐
                    ║ __index   ╫───┤ lookup table ("superclass") │
                    ╚═══╦═══════╝   ├─────────────────────────────┤
  ╔═══════════╗         ║           │ some other method           ┼─── function() ... end
  ║ metatable ║         ║           └─────────────────────────────┘
  ╟───────────╢   ┌─────╨──────────────────┐
  ║ __index   ╫───┤ lookup table ("class") │
  ╚═══╦═══════╝   ├────────────────────────┤
      ║           │ some method            ┼─── function() ... end
      ║           └────────────────────────┘
┌─────╨─────────────────┐
│ base table ("object") │
└───────────────────────┘

Note that a metatable is not the same as a class; it defines behavior, not methods. Conversely, if you try to use a class directly as a metatable, it will probably not do much. (This is pretty different from e.g. Python, where operator overloads are just methods with funny names. One nice thing about the Lua approach is that you can keep interface-like functionality separate from methods, and avoid clogging up arbitrary objects’ namespaces. You could even use a dummy table as a key and completely avoid name collisions.)

Anyway, code!

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
local class = {
    foo = function(a)
        print("foo got", a)
    end,
}
local mt = { __index = class }
-- setmetatable returns its first argument, so this is nice shorthand
local obj1 = setmetatable({}, mt)
local obj2 = setmetatable({}, mt)
obj1.foo(7)  -- foo got 7
obj2.foo(9)  -- foo got 9

Wait, wait, hang on. Didn’t I call these methods? How do they get at the object? Maybe Lua has a magical this variable?

Methods, sort of

Not quite, but this is where the other key feature comes in: method-call syntax. It’s the lightest touch of sugar, just enough to have method invocation.

1
2
3
4
5
6
7
8
9
-- note the colon!
a:b(c, d, ...)

-- exactly equivalent to this
-- (except that `a` is only evaluated once)
a.b(a, c, d, ...)

-- which of course is really this
a["b"](a, c, d, ...)

Now we can write methods that actually do something.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
local class = {
    bar = function(self)
        print("our score is", self.score)
    end,
}
local mt = { __index = class }
local obj1 = setmetatable({ score = 13 }, mt)
local obj2 = setmetatable({ score = 25 }, mt)
obj1:bar()  -- our score is 13
obj2:bar()  -- our score is 25

And that’s all you need. Much like Python, methods and data live in the same namespace, and Lua doesn’t care whether obj:method() finds a function on obj or gets one from the metatable’s __index. Unlike Python, the function will be passed self either way, because self comes from the use of : rather than from the lookup behavior.

(Aside: strictly speaking, any Lua value can have a metatable — and if you try to index a non-table, Lua will always consult the metatable’s __index. Strings all have the string library as a metatable, so you can call methods on them: try ("%s %s"):format(1, 2). I don’t think Lua lets user code set the metatable for non-tables, so this isn’t that interesting, but if you’re writing Lua bindings from C then you can wrap your pointers in metatables to give them methods implemented in C.)

Bringing it all together

Of course, writing all this stuff every time is a little tedious and error-prone, so instead you might want to wrap it all up inside a little function. No problem.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
local function make_object(body)
    -- create a metatable
    local mt = { __index = body }
    -- create a base table to serve as the object itself
    local obj = setmetatable({}, mt)
    -- and, done
    return obj
end

-- you can leave off parens if you're only passing in 
local Dog = {
    -- this acts as a "default" value; if obj.barks is missing, __index will
    -- kick in and find this value on the class.  but if obj.barks is assigned
    -- to, it'll go in the object and shadow the value here.
    barks = 0,

    bark = function(self)
        self.barks = self.barks + 1
        print("woof!")
    end,
}

local mydog = make_object(Dog)
mydog:bark()  -- woof!
mydog:bark()  -- woof!
mydog:bark()  -- woof!
print(mydog.barks)  -- 3
print(Dog.barks)  -- 0

It works, but it’s fairly barebones. The nice thing is that you can extend it pretty much however you want. I won’t reproduce an entire serious object system here — lord knows there are enough of them floating around — but the implementation I have for my LÖVE games lets me do this:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
local Animal = Object:extend{
    cries = 0,
}

-- called automatically by Object
function Animal:init()
    print("whoops i couldn't think of anything interesting to put here")
end

-- this is just nice syntax for adding a first argument called 'self', then
-- assigning this function to Animal.cry
function Animal:cry()
    self.cries = self.cries + 1
end

local Cat = Animal:extend{}

function Cat:cry()
    print("meow!")
    Cat.__super.cry(self)
end

local cat = Cat()
cat:cry()  -- meow!
cat:cry()  -- meow!
print(cat.cries)  -- 2

When I say you can extend it however you want, I mean that. I could’ve implemented Python (2)-style super(Cat, self):cry() syntax; I just never got around to it. I could even make it work with multiple inheritance if I really wanted to — or I could go the complete opposite direction and only implement composition. I could implement descriptors, customizing the behavior of individual table keys. I could add pretty decent syntax for composition/proxying. I am trying very hard to end this section now.

The Lua philosophy

Lua’s philosophy is to… not have a philosophy? It gives you the bare minimum to make objects work, and you can do absolutely whatever you want from there. Lua does have something resembling prototypical inheritance, but it’s not so much a first-class feature as an emergent property of some very simple tools. And since you can make __index be a function, you could avoid the prototypical behavior and do something different entirely.

The very severe downside, of course, is that you have to find or build your own object system — which can get pretty confusing very quickly, what with the multiple small moving parts. Third-party code may also have its own object system with subtly different behavior. (Though, in my experience, third-party code tries very hard to avoid needing an object system at all.)

It’s hard to say what the Lua “culture” is like, since Lua is an embedded language that’s often a little different in each environment. I imagine it has a thousand millicultures, instead. I can say that the tedium of building my own object model has led me into something very “traditional”, with prototypical inheritance and whatnot. It’s partly what I’m used to, but it’s also just really dang easy to get working.

Likewise, while I love properties in Python and use them all the dang time, I’ve yet to use a single one in Lua. They wouldn’t be particularly hard to add to my object model, but having to add them myself (or shop around for an object model with them and also port all my code to use it) adds a huge amount of friction. I’ve thought about designing an interesting ECS with custom object behavior, too, but… is it really worth the effort? For all the power and flexibility Lua offers, the cost is that by the time I have something working at all, I’m too exhausted to actually use any of it.

JavaScript

JavaScript is notable for being preposterously heavily used, yet not having a class block.

Well. Okay. Yes. It has one now. It didn’t for a very long time, and even the one it has now is sugar.

Here’s a vector class again:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
class Vector {
    constructor(x, y) {
        this.x = x;
        this.y = y;
    }

    get magnitude() {
        return Math.sqrt(this.x * this.x + this.y * this.y);
    }

    dot(other) {
        return this.x * other.x + this.y * other.y;
    }
}

In “classic” JavaScript, this would be written as:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
function Vector(x, y) {
    this.x = x;
    this.y = y;
}

Object.defineProperty(Vector.prototype, 'magnitude', {
    configurable: true,
    enumerable: true,
    get: function() {
        return Math.sqrt(this.x * this.x + this.y * this.y);
    },
});


Vector.prototype.dot = function(other) {
    return this.x * other.x + this.y * other.y;
};

Hm, yes. I can see why they added class.

The JavaScript model

In JavaScript, a new type is defined in terms of a function, which is its constructor.

Right away we get into trouble here. There is a very big difference between these two invocations, which I actually completely forgot about just now after spending four hours writing about Python and Lua:

1
2
let vec = Vector(3, 4);
let vec = new Vector(3, 4);

The first calls the function Vector. It assigns some properties to this, which here is going to be window, so now you have a global x and y. It then returns nothing, so vec is undefined.

The second calls Vector with this set to a new empty object, then evaluates to that object. The result is what you’d actually expect.

(You can detect this situation with the strange new.target expression, but I have never once remembered to do so.)

From here, we have true, honest-to-god, first-class prototypical inheritance. The word “prototype” is even right there. When you write this:

1
vec.dot(vec2)

JavaScript will look for dot on vec and (presumably) not find it. It then consults vecs prototype, an object you can see for yourself by using Object.getPrototypeOf(). Since vec is a Vector, its prototype is Vector.prototype.

I stress that Vector.prototype is not the prototype for Vector. It’s the prototype for instances of Vector.

(I say “instance”, but the true type of vec here is still just object. If you want to find Vector, it’s automatically assigned to the constructor property of its own prototype, so it’s available as vec.constructor.)

Of course, Vector.prototype can itself have a prototype, in which case the process would continue if dot were not found. A common (and, arguably, very bad) way to simulate single inheritance is to set Class.prototype to an instance of a superclass to get the prototype right, then tack on the methods for Class. Nowadays we can do Object.create(Superclass.prototype).

Now that I’ve been through Python and Lua, though, this isn’t particularly surprising. I kinda spoiled it.

I suppose one difference in JavaScript is that you can tack arbitrary attributes directly onto Vector all you like, and they will remain invisible to instances since they aren’t in the prototype chain. This is kind of backwards from Lua, where you can squirrel stuff away in the metatable.

Another difference is that every single object in JavaScript has a bunch of properties already tacked on — the ones in Object.prototype. Every object (and by “object” I mean any mapping) has a prototype, and that prototype defaults to Object.prototype, and it has a bunch of ancient junk like isPrototypeOf.

(Nit: it’s possible to explicitly create an object with no prototype via Object.create(null).)

Like Lua, and unlike Python, JavaScript doesn’t distinguish between keys found on an object and keys found via a prototype. Properties can be defined on prototypes with Object.defineProperty(), but that works just as well directly on an object, too. JavaScript doesn’t have a lot of operator overloading, but some things like Symbol.iterator also work on both objects and prototypes.

About this

You may, at this point, be wondering what this is. Unlike Lua and Python (and the last language below), this is a special built-in value — a context value, invisibly passed for every function call.

It’s determined by where the function came from. If the function was the result of an attribute lookup, then this is set to the object containing that attribute. Otherwise, this is set to the global object, window. (You can also set this to whatever you want via the call method on functions.)

This decision is made lexically, i.e. from the literal source code as written. There are no Python-style bound methods. In other words:

1
2
3
4
5
// this = obj
obj.method()
// this = window
let meth = obj.method
meth()

Also, because this is reassigned on every function call, it cannot be meaningfully closed over, which makes using closures within methods incredibly annoying. The old approach was to assign this to some other regular name like self (which got syntax highlighting since it’s also a built-in name in browsers); then we got Function.bind, which produced a callable thing with a fixed context value, which was kind of nice; and now finally we have arrow functions, which explicitly close over the current this when they’re defined and don’t change it when called. Phew.

Class syntax

I already showed class syntax, and it’s really just one big macro for doing all the prototype stuff The Right Way. It even prevents you from calling the type without new. The underlying model is exactly the same, and you can inspect all the parts.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
class Vector { ... }

console.log(Vector.prototype);  // { dot: ..., magnitude: ..., ... }
let vec = new Vector(3, 4);
console.log(Object.getPrototypeOf(vec));  // same as Vector.prototype

// i don't know why you would subclass vector but let's roll with it
class Vectest extends Vector { ... }

console.log(Vectest.prototype);  // { ... }
console.log(Object.getPrototypeOf(Vectest.prototype))  // same as Vector.prototype

Alas, class syntax has a couple shortcomings. You can’t use the class block to assign arbitrary data to either the type object or the prototype — apparently it was deemed too confusing that mutations would be shared among instances. Which… is… how prototypes work. How Python works. How JavaScript itself, one of the most popular languages of all time, has worked for twenty-two years. Argh.

You can still do whatever assignment you want outside of the class block, of course. It’s just a little ugly, and not something I’d think to look for with a sugary class.

A more subtle result of this behavior is that a class block isn’t quite the same syntax as an object literal. The check for data isn’t a runtime thing; class Foo { x: 3 } fails to parse. So JavaScript now has two largely but not entirely identical styles of key/value block.

Attribute access

Here’s where things start to come apart at the seams, just a little bit.

JavaScript doesn’t really have an attribute protocol. Instead, it has two… extension points, I suppose.

One is Object.defineProperty, seen above. For common cases, there’s also the get syntax inside a property literal, which does the same thing. But unlike Python’s @property, these aren’t wrappers around some simple primitives; they are the primitives. JavaScript is the only language of these four to have “property that runs code on access” as a completely separate first-class concept.

If you want to intercept arbitrary attribute access (and some kinds of operators), there’s a completely different primitive: the Proxy type. It doesn’t let you intercept attribute access or operators; instead, it produces a wrapper object that supports interception and defers to the wrapped object by default.

It’s cool to see composition used in this way, but also, extremely weird. If you want to make your own type that overloads in or calling, you have to return a Proxy that wraps your own type, rather than actually returning your own type. And (unlike the other three languages in this post) you can’t return a different type from a constructor, so you have to throw that away and produce objects only from a factory. And instanceof would be broken, but you can at least fix that with Symbol.hasInstance — which is really operator overloading, implement yet another completely different way.

I know the design here is a result of legacy and speed — if any object could intercept all attribute access, then all attribute access would be slowed down everywhere. Fair enough. It still leaves the surface area of the language a bit… bumpy?

The JavaScript philosophy

It’s a little hard to tell. The original idea of prototypes was interesting, but it was hidden behind some very awkward syntax. Since then, we’ve gotten a bunch of extra features awkwardly bolted on to reflect the wildly varied things the built-in types and DOM API were already doing. We have class syntax, but it’s been explicitly designed to avoid exposing the prototype parts of the model.

I admit I don’t do a lot of heavy JavaScript, so I might just be overlooking it, but I’ve seen virtually no code that makes use of any of the recent advances in object capabilities. Forget about custom iterators or overloading call; I can’t remember seeing any JavaScript in the wild that even uses properties yet. I don’t know if everyone’s waiting for sufficient browser support, nobody knows about them, or nobody cares.

The model has advanced recently, but I suspect JavaScript is still shackled to its legacy of “something about prototypes, I don’t really get it, just copy the other code that’s there” as an object model. Alas! Prototypes are so good. Hopefully class syntax will make it a bit more accessible, as it has in Python.

Perl 5

Perl 5 also doesn’t have an object system and expects you to build your own. But where Lua gives you two simple, powerful tools for building one, Perl 5 feels more like a puzzle with half the pieces missing. Clearly they were going for something, but they only gave you half of it.

In brief, a Perl object is a reference that has been blessed with a package.

I need to explain a few things. Honestly, one of the biggest problems with the original Perl object setup was how many strange corners and unique jargon you had to understand just to get off the ground.

(If you want to try running any of this code, you should stick a use v5.26; as the first line. Perl is very big on backwards compatibility, so you need to opt into breaking changes, and even the mundane say builtin is behind a feature gate.)

References

A reference in Perl is sort of like a pointer, but its main use is very different. See, Perl has the strange property that its data structures try very hard to spill their contents all over the place. Despite having dedicated syntax for arrays — @foo is an array variable, distinct from the single scalar variable $foo — it’s actually impossible to nest arrays.

1
2
3
my @foo = (1, 2, 3, 4);
my @bar = (@foo, @foo);
# @bar is now a flat list of eight items: 1, 2, 3, 4, 1, 2, 3, 4

The idea, I guess, is that an array is not one thing. It’s not a container, which happens to hold multiple things; it is multiple things. Anywhere that expects a single value, such as an array element, cannot contain an array, because an array fundamentally is not a single value.

And so we have “references”, which are a form of indirection, but also have the nice property that they’re single values. They add containment around arrays, and in general they make working with most of Perl’s primitive types much more sensible. A reference to a variable can be taken with the \ operator, or you can use [ ... ] and { ... } to directly create references to anonymous arrays or hashes.

1
2
3
my @foo = (1, 2, 3, 4);
my @bar = (\@foo, \@foo);
# @bar is now a nested list of two items: [1, 2, 3, 4], [1, 2, 3, 4]

(Incidentally, this is the sole reason I initially abandoned Perl for Python. Non-trivial software kinda requires nesting a lot of data structures, so you end up with references everywhere, and the syntax for going back and forth between a reference and its contents is tedious and ugly.)

A Perl object must be a reference. Perl doesn’t care what kind of reference — it’s usually a hash reference, since hashes are a convenient place to store arbitrary properties, but it could just as well be a reference to an array, a scalar, or even a sub (i.e. function) or filehandle.

I’m getting a little ahead of myself. First, the other half: blessing and packages.

Packages and blessing

Perl packages are just namespaces. A package looks like this:

1
2
3
4
5
6
7
package Foo::Bar;

sub quux {
    say "hi from quux!";
}

# now Foo::Bar::quux() can be called from anywhere

Nothing shocking, right? It’s just a named container. A lot of the details are kind of weird, like how a package exists in some liminal quasi-value space, but the basic idea is a Bag Of Stuff.

The final piece is “blessing,” which is Perl’s funny name for binding a package to a reference. A very basic class might look like this:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
package Vector;

# the name 'new' is convention, not special
sub new {
    # perl argument passing is weird, don't ask
    my ($class, $x, $y) = @_;

    # create the object itself -- here, unusually, an array reference makes sense
    my $self = [ $x, $y ];

    # associate the package with that reference
    # note that $class here is just the regular string, 'Vector'
    bless $self, $class;

    return $self;
}

sub x {
    my ($self) = @_;
    return $self->[0];
}

sub y {
    my ($self) = @_;
    return $self->[1];
}

sub magnitude {
    my ($self) = @_;
    return sqrt($self->x ** 2 + $self->y ** 2);
}

# switch back to the "default" package
package main;

# -> is method call syntax, which passes the invocant as the first argument;
# for a package, that's just the package name
my $vec = Vector->new(3, 4);
say $vec->magnitude;  # 5

A few things of note here. First, $self->[0] has nothing to do with objects; it’s normal syntax for getting the value of a index 0 out of an array reference called $self. (Most classes are based on hashrefs and would use $self->{value} instead.) A blessed reference is still a reference and can be treated like one.

In general, -> is Perl’s dereferencey operator, but its exact behavior depends on what follows. If it’s followed by brackets, then it’ll apply the brackets to the thing in the reference: ->{} to index a hash reference, ->[] to index an array reference, and ->() to call a function reference.

But if -> is followed by an identifier, then it’s a method call. For packages, that means calling a function in the package and passing the package name as the first argument. For objects — blessed references — that means calling a function in the associated package and passing the object as the first argument.

This is a little weird! A blessed reference is a superposition of two things: its normal reference behavior, and some completely orthogonal object behavior. Also, object behavior has no notion of methods vs data; it only knows about methods. Perl lets you omit parentheses in a lot of places, including when calling a method with no arguments, so $vec->magnitude is really $vec->magnitude().

Perl’s blessing bears some similarities to Lua’s metatables, but ultimately Perl is much closer to Ruby’s “message passing” approach than the above three languages’ approaches of “get me something and maybe it’ll be callable”. (But this is no surprise — Ruby is a spiritual successor to Perl 5.)

All of this leads to one little wrinkle: how do you actually expose data? Above, I had to write x and y methods. Am I supposed to do that for every single attribute on my type?

Yes! But don’t worry, there are third-party modules to help with this incredibly fundamental task. Take Class::Accessor::Fast, so named because it’s faster than Class::Accessor:

1
2
3
package Foo;
use base qw(Class::Accessor::Fast);
__PACKAGE__->mk_accessors(qw(fred wilma barney));

(__PACKAGE__ is the lexical name of the current package; qw(...) is a list literal that splits its contents on whitespace.)

This assumes you’re using a hashref with keys of the same names as the attributes. $obj->fred will return the fred key from your hashref, and $obj->fred(4) will change it to 4.

You also, somewhat bizarrely, have to inherit from Class::Accessor::Fast. Speaking of which,

Inheritance

Inheritance is done by populating the package-global @ISA array with some number of (string) names of parent packages. Most code instead opts to write use base ...;, which does the same thing. Or, more commonly, use parent ...;, which… also… does the same thing.

Every package implicitly inherits from UNIVERSAL, which can be freely modified by Perl code.

A method can call its superclass method with the SUPER:: pseudo-package:

1
2
3
4
sub foo {
    my ($self) = @_;
    $self->SUPER::foo;
}

However, this does a depth-first search, which means it almost certainly does the wrong thing when faced with multiple inheritance. For a while the accepted solution involved a third-party module, but Perl eventually grew an alternative you have to opt into: C3, which may be more familiar to you as the order Python uses.

1
2
3
4
5
6
use mro 'c3';

sub foo {
    my ($self) = @_;
    $self->next::method;
}

Offhand, I’m not actually sure how next::method works, seeing as it was originally implemented in pure Perl code. I suspect it involves peeking at the caller’s stack frame. If so, then this is a very different style of customizability from e.g. Python — the MRO was never intended to be pluggable, and the use of a special pseudo-package means it isn’t really, but someone was determined enough to make it happen anyway.

Operator overloading and whatnot

Operator overloading looks a little weird, though really it’s pretty standard Perl.

1
2
3
4
5
6
7
8
package MyClass;

use overload '+' => \&_add;

sub _add {
    my ($self, $other, $swap) = @_;
    ...
}

use overload here is a pragma, where “pragma” means “regular-ass module that does some wizardry when imported”.

\&_add is how you get a reference to the _add sub so you can pass it to the overload module. If you just said &_add or _add, that would call it.

And that’s it; you just pass a map of operators to functions to this built-in module. No worry about name clashes or pollution, which is pretty nice. You don’t even have to give references to functions that live in the package, if you don’t want them to clog your namespace; you could put them in another package, or even inline them anonymously.

One especially interesting thing is that Perl lets you overload every operator. Perl has a lot of operators. It considers some math builtins like sqrt and trig functions to be operators, or at least operator-y enough that you can overload them. You can also overload the “file text” operators, such as -e $path to test whether a file exists. You can overload conversions, including implicit conversion to a regex. And most fascinating to me, you can overload dereferencing — that is, the thing Perl does when you say $hashref->{key} to get at the underlying hash. So a single object could pretend to be references of multiple different types, including a subref to implement callability. Neat.

Somewhat related: you can overload basic operators (indexing, etc.) on basic types (not references!) with the tie function, which is designed completely differently and looks for methods with fixed names. Go figure.

You can intercept calls to nonexistent methods by implementing a function called AUTOLOAD, within which the $AUTOLOAD global will contain the name of the method being called. Originally this feature was, I think, intended for loading binary components or large libraries on-the-fly only when needed, hence the name. Offhand I’m not sure I ever saw it used the way __getattr__ is used in Python.

Is there a way to intercept all method calls? I don’t think so, but it is Perl, so I must be forgetting something.

Actually no one does this any more

Like a decade ago, a council of elder sages sat down and put together a whole whizbang system that covers all of it: Moose.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
package Vector;
use Moose;

has x => (is => 'rw', isa => 'Int');
has y => (is => 'rw', isa => 'Int');

sub magnitude {
    my ($self) = @_;
    return sqrt($self->x ** 2 + $self->y ** 2);
}

Moose has its own way to do pretty much everything, and it’s all built on the same primitives. Moose also adds metaclasses, somehow, despite that the underlying model doesn’t actually support them? I’m not entirely sure how they managed that, but I do remember doing some class introspection with Moose and it was much nicer than the built-in way.

(If you’re wondering, the built-in way begins with looking at the hash called %Vector::. No, that’s not a typo.)

I really cannot stress enough just how much stuff Moose does, but I don’t want to delve into it here since Moose itself is not actually the language model.

The Perl philosophy

I hope you can see what I meant with what I first said about Perl, now. It has multiple inheritance with an MRO, but uses the wrong one by default. It has extensive operator overloading, which looks nothing like how inheritance works, and also some of it uses a totally different mechanism with special method names instead. It only understands methods, not data, leaving you to figure out accessors by hand.

There’s 70% of an object system here with a clear general design it was gunning for, but none of the pieces really look anything like each other. It’s weird, in a distinctly Perl way.

The result is certainly flexible, at least! It’s especially cool that you can use whatever kind of reference you want for storage, though even as I say that, I acknowledge it’s no different from simply subclassing list or something in Python. It feels different in Perl, but maybe only because it looks so different.

I haven’t written much Perl in a long time, so I don’t know what the community is like any more. Moose was already ubiquitous when I left, which you’d think would let me say “the community mostly focuses on the stuff Moose can do” — but even a decade ago, Moose could already do far more than I had ever seen done by hand in Perl. It’s always made a big deal out of roles (read: interfaces), for instance, despite that I’d never seen anyone care about them in Perl before Moose came along. Maybe their presence in Moose has made them more popular? Who knows.

Also, I wrote Perl seriously, but in the intervening years I’ve only encountered people who only ever used Perl for one-offs. Maybe it’ll come as a surprise to a lot of readers that Perl has an object model at all.

End

Well, that was fun! I hope any of that made sense.

Special mention goes to Rust, which doesn’t have an object model you can fiddle with at runtime, but does do things a little differently.

It’s been really interesting thinking about how tiny differences make a huge impact on what people do in practice. Take the choice of storage in Perl versus Python. Perl’s massively common URI class uses a string as the storage, nothing else; I haven’t seen anything like that in Python aside from markupsafe, which is specifically designed as a string type. I would guess this is partly because Perl makes you choose — using a hashref is an obvious default, but you have to make that choice one way or the other. In Python (especially 3), inheriting from object and getting dict-based storage is the obvious thing to do; the ability to use another type isn’t quite so obvious, and doing it “right” involves a tiny bit of extra work.

Or, consider that Lua could have descriptors, but the extra bit of work (especially design work) has been enough of an impediment that I’ve never implemented them. I don’t think the object implementations I’ve looked at have included them, either. Super weird!

In that light, it’s only natural that objects would be so strongly associated with the features Java and C++ attach to them. I think that makes it all the more important to play around! Look at what Moose has done. No, really, you should bear in mind my description of how Perl does stuff and flip through the Moose documentation. It’s amazing what they’ve built.

Using Amazon Redshift Spectrum, Amazon Athena, and AWS Glue with Node.js in Production

Post Syndicated from Rafi Ton original https://aws.amazon.com/blogs/big-data/using-amazon-redshift-spectrum-amazon-athena-and-aws-glue-with-node-js-in-production/

This is a guest post by Rafi Ton, founder and CEO of NUVIAD. NUVIAD is, in their own words, “a mobile marketing platform providing professional marketers, agencies and local businesses state of the art tools to promote their products and services through hyper targeting, big data analytics and advanced machine learning tools.”

At NUVIAD, we’ve been using Amazon Redshift as our main data warehouse solution for more than 3 years.

We store massive amounts of ad transaction data that our users and partners analyze to determine ad campaign strategies. When running real-time bidding (RTB) campaigns in large scale, data freshness is critical so that our users can respond rapidly to changes in campaign performance. We chose Amazon Redshift because of its simplicity, scalability, performance, and ability to load new data in near real time.

Over the past three years, our customer base grew significantly and so did our data. We saw our Amazon Redshift cluster grow from three nodes to 65 nodes. To balance cost and analytics performance, we looked for a way to store large amounts of less-frequently analyzed data at a lower cost. Yet, we still wanted to have the data immediately available for user queries and to meet their expectations for fast performance. We turned to Amazon Redshift Spectrum.

In this post, I explain the reasons why we extended Amazon Redshift with Redshift Spectrum as our modern data warehouse. I cover how our data growth and the need to balance cost and performance led us to adopt Redshift Spectrum. I also share key performance metrics in our environment, and discuss the additional AWS services that provide a scalable and fast environment, with data available for immediate querying by our growing user base.

Amazon Redshift as our foundation

The ability to provide fresh, up-to-the-minute data to our customers and partners was always a main goal with our platform. We saw other solutions provide data that was a few hours old, but this was not good enough for us. We insisted on providing the freshest data possible. For us, that meant loading Amazon Redshift in frequent micro batches and allowing our customers to query Amazon Redshift directly to get results in near real time.

The benefits were immediately evident. Our customers could see how their campaigns performed faster than with other solutions, and react sooner to the ever-changing media supply pricing and availability. They were very happy.

However, this approach required Amazon Redshift to store a lot of data for long periods, and our data grew substantially. In our peak, we maintained a cluster running 65 DC1.large nodes. The impact on our Amazon Redshift cluster was evident, and we saw our CPU utilization grow to 90%.

Why we extended Amazon Redshift to Redshift Spectrum

Redshift Spectrum gives us the ability to run SQL queries using the powerful Amazon Redshift query engine against data stored in Amazon S3, without needing to load the data. With Redshift Spectrum, we store data where we want, at the cost that we want. We have the data available for analytics when our users need it with the performance they expect.

Seamless scalability, high performance, and unlimited concurrency

Scaling Redshift Spectrum is a simple process. First, it allows us to leverage Amazon S3 as the storage engine and get practically unlimited data capacity.

Second, if we need more compute power, we can leverage Redshift Spectrum’s distributed compute engine over thousands of nodes to provide superior performance – perfect for complex queries running against massive amounts of data.

Third, all Redshift Spectrum clusters access the same data catalog so that we don’t have to worry about data migration at all, making scaling effortless and seamless.

Lastly, since Redshift Spectrum distributes queries across potentially thousands of nodes, they are not affected by other queries, providing much more stable performance and unlimited concurrency.

Keeping it SQL

Redshift Spectrum uses the same query engine as Amazon Redshift. This means that we did not need to change our BI tools or query syntax, whether we used complex queries across a single table or joins across multiple tables.

An interesting capability introduced recently is the ability to create a view that spans both Amazon Redshift and Redshift Spectrum external tables. With this feature, you can query frequently accessed data in your Amazon Redshift cluster and less-frequently accessed data in Amazon S3, using a single view.

Leveraging Parquet for higher performance

Parquet is a columnar data format that provides superior performance and allows Redshift Spectrum (or Amazon Athena) to scan significantly less data. With less I/O, queries run faster and we pay less per query. You can read all about Parquet at https://parquet.apache.org/ or https://en.wikipedia.org/wiki/Apache_Parquet.

Lower cost

From a cost perspective, we pay standard rates for our data in Amazon S3, and only small amounts per query to analyze data with Redshift Spectrum. Using the Parquet format, we can significantly reduce the amount of data scanned. Our costs are now lower, and our users get fast results even for large complex queries.

What we learned about Amazon Redshift vs. Redshift Spectrum performance

When we first started looking at Redshift Spectrum, we wanted to put it to the test. We wanted to know how it would compare to Amazon Redshift, so we looked at two key questions:

  1. What is the performance difference between Amazon Redshift and Redshift Spectrum on simple and complex queries?
  2. Does the data format impact performance?

During the migration phase, we had our dataset stored in Amazon Redshift and S3 as CSV/GZIP and as Parquet file formats. We tested three configurations:

  • Amazon Redshift cluster with 28 DC1.large nodes
  • Redshift Spectrum using CSV/GZIP
  • Redshift Spectrum using Parquet

We performed benchmarks for simple and complex queries on one month’s worth of data. We tested how much time it took to perform the query, and how consistent the results were when running the same query multiple times. The data we used for the tests was already partitioned by date and hour. Properly partitioning the data improves performance significantly and reduces query times.

Simple query

First, we tested a simple query aggregating billing data across a month:

SELECT 
  user_id, 
  count(*) AS impressions, 
  SUM(billing)::decimal /1000000 AS billing 
FROM <table_name> 
WHERE 
  date >= '2017-08-01' AND 
  date <= '2017-08-31'  
GROUP BY 
  user_id;

We ran the same query seven times and measured the response times (red marking the longest time and green the shortest time):

Execution Time (seconds)
  Amazon Redshift Redshift Spectrum
CSV
Redshift Spectrum Parquet
Run #1 39.65 45.11 11.92
Run #2 15.26 43.13 12.05
Run #3 15.27 46.47 13.38
Run #4 21.22 51.02 12.74
Run #5 17.27 43.35 11.76
Run #6 16.67 44.23 13.67
Run #7 25.37 40.39 12.75
Average 21.53  44.82 12.61

For simple queries, Amazon Redshift performed better than Redshift Spectrum, as we thought, because the data is local to Amazon Redshift.

What was surprising was that using Parquet data format in Redshift Spectrum significantly beat ‘traditional’ Amazon Redshift performance. For our queries, using Parquet data format with Redshift Spectrum delivered an average 40% performance gain over traditional Amazon Redshift. Furthermore, Redshift Spectrum showed high consistency in execution time with a smaller difference between the slowest run and the fastest run.

Comparing the amount of data scanned when using CSV/GZIP and Parquet, the difference was also significant:

Data Scanned (GB)
CSV (Gzip) 135.49
Parquet 2.83

Because we pay only for the data scanned by Redshift Spectrum, the cost saving of using Parquet is evident and substantial.

Complex query

Next, we compared the same three configurations with a complex query.

Execution Time (seconds)
  Amazon Redshift Redshift Spectrum CSV Redshift Spectrum Parquet
Run #1 329.80 84.20 42.40
Run #2 167.60 65.30 35.10
Run #3 165.20 62.20 23.90
Run #4 273.90 74.90 55.90
Run #5 167.70 69.00 58.40
Average 220.84 71.12 43.14

This time, Redshift Spectrum using Parquet cut the average query time by 80% compared to traditional Amazon Redshift!

Bottom line: For complex queries, Redshift Spectrum provided a 67% performance gain over Amazon Redshift. Using the Parquet data format, Redshift Spectrum delivered an 80% performance improvement over Amazon Redshift. For us, this was substantial.

Optimizing the data structure for different workloads

Because the cost of S3 is relatively inexpensive and we pay only for the data scanned by each query, we believe that it makes sense to keep our data in different formats for different workloads and different analytics engines. It is important to note that we can have any number of tables pointing to the same data on S3. It all depends on how we partition the data and update the table partitions.

Data permutations

For example, we have a process that runs every minute and generates statistics for the last minute of data collected. With Amazon Redshift, this would be done by running the query on the table with something as follows:

SELECT 
  user, 
  COUNT(*) 
FROM 
  events_table 
WHERE 
  ts BETWEEN ‘2017-08-01 14:00:00’ AND ‘2017-08-01 14:00:59’ 
GROUP BY 
  user;

(Assuming ‘ts’ is your column storing the time stamp for each event.)

With Redshift Spectrum, we pay for the data scanned in each query. If the data is partitioned by the minute instead of the hour, a query looking at one minute would be 1/60th the cost. If we use a temporary table that points only to the data of the last minute, we save that unnecessary cost.

Creating Parquet data efficiently

On the average, we have 800 instances that process our traffic. Each instance sends events that are eventually loaded into Amazon Redshift. When we started three years ago, we would offload data from each server to S3 and then perform a periodic copy command from S3 to Amazon Redshift.

Recently, Amazon Kinesis Firehose added the capability to offload data directly to Amazon Redshift. While this is now a viable option, we kept the same collection process that worked flawlessly and efficiently for three years.

This changed, however, when we incorporated Redshift Spectrum. With Redshift Spectrum, we needed to find a way to:

  • Collect the event data from the instances.
  • Save the data in Parquet format.
  • Partition the data effectively.

To accomplish this, we save the data as CSV and then transform it to Parquet. The most effective method to generate the Parquet files is to:

  1. Send the data in one-minute intervals from the instances to Kinesis Firehose with an S3 temporary bucket as the destination.
  2. Aggregate hourly data and convert it to Parquet using AWS Lambda and AWS Glue.
  3. Add the Parquet data to S3 by updating the table partitions.

With this new process, we had to give more attention to validating the data before we sent it to Kinesis Firehose, because a single corrupted record in a partition fails queries on that partition.

Data validation

To store our click data in a table, we considered the following SQL create table command:

create external TABLE spectrum.blog_clicks (
    user_id varchar(50),
    campaign_id varchar(50),
    os varchar(50),
    ua varchar(255),
    ts bigint,
    billing float
)
partitioned by (date date, hour smallint)  
stored as parquet
location 's3://nuviad-temp/blog/clicks/';

The above statement defines a new external table (all Redshift Spectrum tables are external tables) with a few attributes. We stored ‘ts’ as a Unix time stamp and not as Timestamp, and billing data is stored as float and not decimal (more on that later). We also said that the data is partitioned by date and hour, and then stored as Parquet on S3.

First, we need to get the table definitions. This can be achieved by running the following query:

SELECT 
  * 
FROM 
  svv_external_columns 
WHERE 
  tablename = 'blog_clicks';

This query lists all the columns in the table with their respective definitions:

schemaname tablename columnname external_type columnnum part_key
spectrum blog_clicks user_id varchar(50) 1 0
spectrum blog_clicks campaign_id varchar(50) 2 0
spectrum blog_clicks os varchar(50) 3 0
spectrum blog_clicks ua varchar(255) 4 0
spectrum blog_clicks ts bigint 5 0
spectrum blog_clicks billing double 6 0
spectrum blog_clicks date date 7 1
spectrum blog_clicks hour smallint 8 2

Now we can use this data to create a validation schema for our data:

const rtb_request_schema = {
    "name": "clicks",
    "items": {
        "user_id": {
            "type": "string",
            "max_length": 100
        },
        "campaign_id": {
            "type": "string",
            "max_length": 50
        },
        "os": {
            "type": "string",
            "max_length": 50            
        },
        "ua": {
            "type": "string",
            "max_length": 255            
        },
        "ts": {
            "type": "integer",
            "min_value": 0,
            "max_value": 9999999999999
        },
        "billing": {
            "type": "float",
            "min_value": 0,
            "max_value": 9999999999999
        }
    }
};

Next, we create a function that uses this schema to validate data:

function valueIsValid(value, item_schema) {
    if (schema.type == 'string') {
        return (typeof value == 'string' && value.length <= schema.max_length);
    }
    else if (schema.type == 'integer') {
        return (typeof value == 'number' && value >= schema.min_value && value <= schema.max_value);
    }
    else if (schema.type == 'float' || schema.type == 'double') {
        return (typeof value == 'number' && value >= schema.min_value && value <= schema.max_value);
    }
    else if (schema.type == 'boolean') {
        return typeof value == 'boolean';
    }
    else if (schema.type == 'timestamp') {
        return (new Date(value)).getTime() > 0;
    }
    else {
        return true;
    }
}

Near real-time data loading with Kinesis Firehose

On Kinesis Firehose, we created a new delivery stream to handle the events as follows:

Delivery stream name: events
Source: Direct PUT
S3 bucket: nuviad-events
S3 prefix: rtb/
IAM role: firehose_delivery_role_1
Data transformation: Disabled
Source record backup: Disabled
S3 buffer size (MB): 100
S3 buffer interval (sec): 60
S3 Compression: GZIP
S3 Encryption: No Encryption
Status: ACTIVE
Error logging: Enabled

This delivery stream aggregates event data every minute, or up to 100 MB, and writes the data to an S3 bucket as a CSV/GZIP compressed file. Next, after we have the data validated, we can safely send it to our Kinesis Firehose API:

if (validated) {
    let itemString = item.join('|')+'\n'; //Sending csv delimited by pipe and adding new line

    let params = {
        DeliveryStreamName: 'events',
        Record: {
            Data: itemString
        }
    };

    firehose.putRecord(params, function(err, data) {
        if (err) {
            console.error(err, err.stack);        
        }
        else {
            // Continue to your next step 
        }
    });
}

Now, we have a single CSV file representing one minute of event data stored in S3. The files are named automatically by Kinesis Firehose by adding a UTC time prefix in the format YYYY/MM/DD/HH before writing objects to S3. Because we use the date and hour as partitions, we need to change the file naming and location to fit our Redshift Spectrum schema.

Automating data distribution using AWS Lambda

We created a simple Lambda function triggered by an S3 put event that copies the file to a different location (or locations), while renaming it to fit our data structure and processing flow. As mentioned before, the files generated by Kinesis Firehose are structured in a pre-defined hierarchy, such as:

S3://your-bucket/your-prefix/2017/08/01/20/events-4-2017-08-01-20-06-06-536f5c40-6893-4ee4-907d-81e4d3b09455.gz

All we need to do is parse the object name and restructure it as we see fit. In our case, we did the following (the event is an object received in the Lambda function with all the data about the object written to S3):

/*
	object key structure in the event object:
your-prefix/2017/08/01/20/event-4-2017-08-01-20-06-06-536f5c40-6893-4ee4-907d-81e4d3b09455.gz
	*/

let key_parts = event.Records[0].s3.object.key.split('/'); 

let event_type = key_parts[0];
let date = key_parts[1] + '-' + key_parts[2] + '-' + key_parts[3];
let hour = key_parts[4];
if (hour.indexOf('0') == 0) {
 		hour = parseInt(hour, 10) + '';
}
    
let parts1 = key_parts[5].split('-');
let minute = parts1[7];
if (minute.indexOf('0') == 0) {
        minute = parseInt(minute, 10) + '';
}

Now, we can redistribute the file to the two destinations we need—one for the minute processing task and the other for hourly aggregation:

    copyObjectToHourlyFolder(event, date, hour, minute)
        .then(copyObjectToMinuteFolder.bind(null, event, date, hour, minute))
        .then(addPartitionToSpectrum.bind(null, event, date, hour, minute))
        .then(deleteOldMinuteObjects.bind(null, event))
        .then(deleteStreamObject.bind(null, event))        
        .then(result => {
            callback(null, { message: 'done' });            
        })
        .catch(err => {
            console.error(err);
            callback(null, { message: err });            
        }); 

Kinesis Firehose stores the data in a temporary folder. We copy the object to another folder that holds the data for the last processed minute. This folder is connected to a small Redshift Spectrum table where the data is being processed without needing to scan a much larger dataset. We also copy the data to a folder that holds the data for the entire hour, to be later aggregated and converted to Parquet.

Because we partition the data by date and hour, we created a new partition on the Redshift Spectrum table if the processed minute is the first minute in the hour (that is, minute 0). We ran the following:

ALTER TABLE 
  spectrum.events 
ADD partition
  (date='2017-08-01', hour=0) 
  LOCATION 's3://nuviad-temp/events/2017-08-01/0/';

After the data is processed and added to the table, we delete the processed data from the temporary Kinesis Firehose storage and from the minute storage folder.

Migrating CSV to Parquet using AWS Glue and Amazon EMR

The simplest way we found to run an hourly job converting our CSV data to Parquet is using Lambda and AWS Glue (and thanks to the awesome AWS Big Data team for their help with this).

Creating AWS Glue jobs

What this simple AWS Glue script does:

  • Gets parameters for the job, date, and hour to be processed
  • Creates a Spark EMR context allowing us to run Spark code
  • Reads CSV data into a DataFrame
  • Writes the data as Parquet to the destination S3 bucket
  • Adds or modifies the Redshift Spectrum / Amazon Athena table partition for the table
import sys
import sys
from awsglue.transforms import *
from awsglue.utils import getResolvedOptions
from pyspark.context import SparkContext
from awsglue.context import GlueContext
from awsglue.job import Job
import boto3

## @params: [JOB_NAME]
args = getResolvedOptions(sys.argv, ['JOB_NAME','day_partition_key', 'hour_partition_key', 'day_partition_value', 'hour_partition_value' ])

#day_partition_key = "partition_0"
#hour_partition_key = "partition_1"
#day_partition_value = "2017-08-01"
#hour_partition_value = "0"

day_partition_key = args['day_partition_key']
hour_partition_key = args['hour_partition_key']
day_partition_value = args['day_partition_value']
hour_partition_value = args['hour_partition_value']

print("Running for " + day_partition_value + "/" + hour_partition_value)

sc = SparkContext()
glueContext = GlueContext(sc)
spark = glueContext.spark_session
job = Job(glueContext)
job.init(args['JOB_NAME'], args)

df = spark.read.option("delimiter","|").csv("s3://nuviad-temp/events/"+day_partition_value+"/"+hour_partition_value)
df.registerTempTable("data")

df1 = spark.sql("select _c0 as user_id, _c1 as campaign_id, _c2 as os, _c3 as ua, cast(_c4 as bigint) as ts, cast(_c5 as double) as billing from data")

df1.repartition(1).write.mode("overwrite").parquet("s3://nuviad-temp/parquet/"+day_partition_value+"/hour="+hour_partition_value)

client = boto3.client('athena', region_name='us-east-1')

response = client.start_query_execution(
    QueryString='alter table parquet_events add if not exists partition(' + day_partition_key + '=\'' + day_partition_value + '\',' + hour_partition_key + '=' + hour_partition_value + ')  location \'s3://nuviad-temp/parquet/' + day_partition_value + '/hour=' + hour_partition_value + '\'' ,
    QueryExecutionContext={
        'Database': 'spectrumdb'
    },
    ResultConfiguration={
        'OutputLocation': 's3://nuviad-temp/convertresults'
    }
)

response = client.start_query_execution(
    QueryString='alter table parquet_events partition(' + day_partition_key + '=\'' + day_partition_value + '\',' + hour_partition_key + '=' + hour_partition_value + ') set location \'s3://nuviad-temp/parquet/' + day_partition_value + '/hour=' + hour_partition_value + '\'' ,
    QueryExecutionContext={
        'Database': 'spectrumdb'
    },
    ResultConfiguration={
        'OutputLocation': 's3://nuviad-temp/convertresults'
    }
)

job.commit()

Note: Because Redshift Spectrum and Athena both use the AWS Glue Data Catalog, we could use the Athena client to add the partition to the table.

Here are a few words about float, decimal, and double. Using decimal proved to be more challenging than we expected, as it seems that Redshift Spectrum and Spark use them differently. Whenever we used decimal in Redshift Spectrum and in Spark, we kept getting errors, such as:

S3 Query Exception (Fetch). Task failed due to an internal error. File 'https://s3-external-1.amazonaws.com/nuviad-temp/events/2017-08-01/hour=2/part-00017-48ae5b6b-906e-4875-8cde-bc36c0c6d0ca.c000.snappy.parquet has an incompatible Parquet schema for column 's3://nuviad-events/events.lat'. Column type: DECIMAL(18, 8), Parquet schema:\noptional float lat [i:4 d:1 r:0]\n (https://s3-external-1.amazonaws.com/nuviad-temp/events/2017-08-01/hour=2/part-00017-48ae5b6b-906e-4875-8cde-bc36c0c6d0ca.c000.snappy.parq

We had to experiment with a few floating-point formats until we found that the only combination that worked was to define the column as double in the Spark code and float in Spectrum. This is the reason you see billing defined as float in Spectrum and double in the Spark code.

Creating a Lambda function to trigger conversion

Next, we created a simple Lambda function to trigger the AWS Glue script hourly using a simple Python code:

import boto3
import json
from datetime import datetime, timedelta
 
client = boto3.client('glue')
 
def lambda_handler(event, context):
    last_hour_date_time = datetime.now() - timedelta(hours = 1)
    day_partition_value = last_hour_date_time.strftime("%Y-%m-%d") 
    hour_partition_value = last_hour_date_time.strftime("%-H") 
    response = client.start_job_run(
    JobName='convertEventsParquetHourly',
    Arguments={
         '--day_partition_key': 'date',
         '--hour_partition_key': 'hour',
         '--day_partition_value': day_partition_value,
         '--hour_partition_value': hour_partition_value
         }
    )

Using Amazon CloudWatch Events, we trigger this function hourly. This function triggers an AWS Glue job named ‘convertEventsParquetHourly’ and runs it for the previous hour, passing job names and values of the partitions to process to AWS Glue.

Redshift Spectrum and Node.js

Our development stack is based on Node.js, which is well-suited for high-speed, light servers that need to process a huge number of transactions. However, a few limitations of the Node.js environment required us to create workarounds and use other tools to complete the process.

Node.js and Parquet

The lack of Parquet modules for Node.js required us to implement an AWS Glue/Amazon EMR process to effectively migrate data from CSV to Parquet. We would rather save directly to Parquet, but we couldn’t find an effective way to do it.

One interesting project in the works is the development of a Parquet NPM by Marc Vertes called node-parquet (https://www.npmjs.com/package/node-parquet). It is not in a production state yet, but we think it would be well worth following the progress of this package.

Timestamp data type

According to the Parquet documentation, Timestamp data are stored in Parquet as 64-bit integers. However, JavaScript does not support 64-bit integers, because the native number type is a 64-bit double, giving only 53 bits of integer range.

The result is that you cannot store Timestamp correctly in Parquet using Node.js. The solution is to store Timestamp as string and cast the type to Timestamp in the query. Using this method, we did not witness any performance degradation whatsoever.

Lessons learned

You can benefit from our trial-and-error experience.

Lesson #1: Data validation is critical

As mentioned earlier, a single corrupt entry in a partition can fail queries running against this partition, especially when using Parquet, which is harder to edit than a simple CSV file. Make sure that you validate your data before scanning it with Redshift Spectrum.

Lesson #2: Structure and partition data effectively

One of the biggest benefits of using Redshift Spectrum (or Athena for that matter) is that you don’t need to keep nodes up and running all the time. You pay only for the queries you perform and only for the data scanned per query.

Keeping different permutations of your data for different queries makes a lot of sense in this case. For example, you can partition your data by date and hour to run time-based queries, and also have another set partitioned by user_id and date to run user-based queries. This results in faster and more efficient performance of your data warehouse.

Storing data in the right format

Use Parquet whenever you can. The benefits of Parquet are substantial. Faster performance, less data to scan, and much more efficient columnar format. However, it is not supported out-of-the-box by Kinesis Firehose, so you need to implement your own ETL. AWS Glue is a great option.

Creating small tables for frequent tasks

When we started using Redshift Spectrum, we saw our Amazon Redshift costs jump by hundreds of dollars per day. Then we realized that we were unnecessarily scanning a full day’s worth of data every minute. Take advantage of the ability to define multiple tables on the same S3 bucket or folder, and create temporary and small tables for frequent queries.

Lesson #3: Combine Athena and Redshift Spectrum for optimal performance

Moving to Redshift Spectrum also allowed us to take advantage of Athena as both use the AWS Glue Data Catalog. Run fast and simple queries using Athena while taking advantage of the advanced Amazon Redshift query engine for complex queries using Redshift Spectrum.

Redshift Spectrum excels when running complex queries. It can push many compute-intensive tasks, such as predicate filtering and aggregation, down to the Redshift Spectrum layer, so that queries use much less of your cluster’s processing capacity.

Lesson #4: Sort your Parquet data within the partition

We achieved another performance improvement by sorting data within the partition using sortWithinPartitions(sort_field). For example:

df.repartition(1).sortWithinPartitions("campaign_id")…

Conclusion

We were extremely pleased with using Amazon Redshift as our core data warehouse for over three years. But as our client base and volume of data grew substantially, we extended Amazon Redshift to take advantage of scalability, performance, and cost with Redshift Spectrum.

Redshift Spectrum lets us scale to virtually unlimited storage, scale compute transparently, and deliver super-fast results for our users. With Redshift Spectrum, we store data where we want at the cost we want, and have the data available for analytics when our users need it with the performance they expect.


About the Author

With 7 years of experience in the AdTech industry and 15 years in leading technology companies, Rafi Ton is the founder and CEO of NUVIAD. He enjoys exploring new technologies and putting them to use in cutting edge products and services, in the real world generating real money. Being an experienced entrepreneur, Rafi believes in practical-programming and fast adaptation of new technologies to achieve a significant market advantage.

 

 

AWS Media Services – Process, Store, and Monetize Cloud-Based Video

Post Syndicated from Jeff Barr original https://aws.amazon.com/blogs/aws/aws-media-services-process-store-and-monetize-cloud-based-video/

Do you remember what web video was like in the early days? Standalone players, video no larger than a postage stamp, slow & cantankerous connections, overloaded servers, and the ever-present buffering messages were the norm less than two decades ago.

Today, thanks to technological progress and a broad array of standards, things are a lot better. Video consumers are now in control. They use devices of all shapes, sizes, and vintages to enjoy live and recorded content that is broadcast, streamed, or sent over-the-top (OTT, as they say), and expect immediate access to content that captures and then holds their attention. Meeting these expectations presents a challenge for content creators and distributors. Instead of generating video in a one-size-fits-all format, they (or their media servers) must be prepared to produce video that spans a broad range of sizes, formats, and bit rates, taking care to be ready to deal with planned or unplanned surges in demand. In the face of all of this complexity, they must backstop their content with a monetization model that supports the content and the infrastructure to deliver it.

New AWS Media Services
Today we are launching an array of broadcast-quality media services, each designed to address one or more aspects of the challenge that I outlined above. You can use them together to build a complete end-to-end video solution or you can use one or more in building-block style. In true AWS fashion, you can spend more time innovating and less time setting up and running infrastructure, leaving you ready to focus on creating, delivering, and monetizing your content. The services are all elastic, allowing you to ramp up processing power, connections, and storage and giving you the ability to handle million-user (and beyond) spikes with ease.

Here are the services (all accessible from a set of interactive consoles as well as through a comprehensive set of APIs):

AWS Elemental MediaConvert – File-based transcoding for OTT, broadcast, or archiving, with support for a long list of formats and codecs. Features include multi-channel audio, graphic overlays, closed captioning, and several DRM options.

AWS Elemental MediaLive – Live encoding to deliver video streams in real time to both televisions and multiscreen devices. Allows you to deploy highly reliable live channels in minutes, with full control over encoding parameters. It supports ad insertion, multi-channel audio, graphic overlays, and closed captioning.

AWS Elemental MediaPackage – Video origination and just-in-time packaging. Starting from a single input, produces output for multiple devices representing a long list of current and legacy formats. Supports multiple monetization models, time-shifted live streaming, ad insertion, DRM, and blackout management.

AWS Elemental MediaStore – Media-optimized storage that enables high performance and low latency applications such as live streaming, while taking advantage of the scale and durability of Amazon Simple Storage Service (S3).

AWS Elemental MediaTailor – Monetization service that supports ad serving and server-side ad insertion, a broad range of devices, transcoding, and accurate reporting of server-side and client-side ad insertion.

Instead of listing out all of the features in the sections below, I’ve simply included as many screen shots as possible with the expectation that this will give you a better sense of the rich set of features, parameters, and settings that you get with this set of services.

AWS Elemental MediaConvert
MediaConvert allows you to transcode content that is stored in files. You can process individual files or entire media libraries, or anything in-between. You simply create a conversion job that specifies the content and the desired outputs, and submit it to MediaConvert. There’s no software to install or patch and the service scales to meet your needs without affecting turnaround time or performance.

The MediaConvert Console lets you manage Output presets, Job templates, Queues, and Jobs:

You can use a built-in system preset or you can make one of your own. You have full control over the settings when you make your own:

Jobs templates are named, and produce one or more output groups. You can add a new group to a template with a click:

When everything is ready to go, you create a job and make some final selections, then click on Create:

Each account starts with a default queue for jobs, where incoming work is processed in parallel using all processing resources available to the account. Adding queues does not add processing resources, but does cause them to be apportioned across queues. You can temporarily pause one queue in order to devote more resources to the others. You can submit jobs to paused queues and you can also cancel any that have yet to start.

Pricing for this service is based on the amount of video that you process and the features that you use.

AWS Elemental MediaLive
This service is for live encoding, and can be run 24×7. MediaLive channels are deployed on redundant resources distributed in two physically separated Availability Zones in order to provide the reliability expected by our customers in the broadcast industry. You can specify your inputs and define your channels in the MediaLive Console:

After you create an Input, you create a Channel and attach it to the Input:

You have full control over the settings for each channel:

 

AWS Elemental MediaPackage
This service lets you deliver video to many devices from a single source. It focuses on protection and just-in-time packaging, giving you the ability to provide your users with the desired content on the device of their choice. You simply create a channel to get started:

Then you add one or more endpoints. Once again, plenty of options and full control, including a startover window and a time delay:

You find the input URL, user name, and password for your channel and route your live video stream to it for packaging:

AWS Elemental MediaStore
MediaStore offers the performance, consistency, and latency required for live and on-demand media delivery. Objects are written and read into a new “temporal” tier of object storage for a limited amount of time, then move silently into S3 for long-lived durability. You simply create a storage container to group your media content:

The container is available within a minute or so:

Like S3 buckets, MediaStore containers have access policies and no limits on the number of objects or storage capacity.

MediaStore helps you to take full advantage of S3 by managing the object key names so as to maximize storage and retrieval throughput, in accord with the Request Rate and Performance Considerations.

AWS Elemental MediaTailor
This service takes care of server-side ad insertion while providing a broadcast-quality viewer experience by transcoding ad assets on the fly. Your customer’s video player asks MediaTailor for a playlist. MediaTailor, in turn, calls your Ad Decision Server and returns a playlist that references the origin server for your original video and the ads recommended by the Ad Decision Server. The video player makes all of its requests to a single endpoint in order to ensure that client-side ad-blocking is ineffective. You simply create a MediaTailor Configuration:

Context information is passed to the Ad Decision Server in the URL:

Despite the length of this post I have barely scratched the surface of the AWS Media Services. Once AWS re:Invent is in the rear view mirror I hope to do a deep dive and show you how to use each of these services.

Available Now
The entire set of AWS Media Services is available now and you can start using them today! Pricing varies by service, but is built around a pay-as-you-go model.

Jeff;

Bringing Datacenter-Scale Hardware-Software Co-design to the Cloud with FireSim and Amazon EC2 F1 Instances

Post Syndicated from Mia Champion original https://aws.amazon.com/blogs/compute/bringing-datacenter-scale-hardware-software-co-design-to-the-cloud-with-firesim-and-amazon-ec2-f1-instances/

The recent addition of Xilinx FPGAs to AWS Cloud compute offerings is one way that AWS is enabling global growth in the areas of advanced analytics, deep learning and AI. The customized F1 servers use pooled accelerators, enabling interconnectivity of up to 8 FPGAs, each one including 64 GiB DDR4 ECC protected memory, with a dedicated PCIe x16 connection. That makes this a powerful engine with the capacity to process advanced analytical applications at scale, at a significantly faster rate. For example, AWS commercial partner Edico Genome is able to achieve an approximately 30X speedup in analyzing whole genome sequencing datasets using their DRAGEN platform powered with F1 instances.

While the availability of FPGA F1 compute on-demand provides clear accessibility and cost advantages, many mainstream users are still finding that the “threshold to entry” in developing or running FPGA-accelerated simulations is too high. Researchers at the UC Berkeley RISE Lab have developed “FireSim”, powered by Amazon FPGA F1 instances as an open-source resource, FireSim lowers that entry bar and makes it easier for everyone to leverage the power of an FPGA-accelerated compute environment. Whether you are part of a small start-up development team or working at a large datacenter scale, hardware-software co-design enables faster time-to-deployment, lower costs, and more predictable performance. We are excited to feature FireSim in this post from Sagar Karandikar and his colleagues at UC-Berkeley.

―Mia Champion, Sr. Data Scientist, AWS

Mapping an 8-node FireSim cluster simulation to Amazon EC2 F1

As traditional hardware scaling nears its end, the data centers of tomorrow are trending towards heterogeneity, employing custom hardware accelerators and increasingly high-performance interconnects. Prototyping new hardware at scale has traditionally been either extremely expensive, or very slow. In this post, I introduce FireSim, a new hardware simulation platform under development in the computer architecture research group at UC Berkeley that enables fast, scalable hardware simulation using Amazon EC2 F1 instances.

FireSim benefits both hardware and software developers working on new rack-scale systems: software developers can use the simulated nodes with new hardware features as they would use a real machine, while hardware developers have full control over the hardware being simulated and can run real software stacks while hardware is still under development. In conjunction with this post, we’re releasing the first public demo of FireSim, which lets you deploy your own 8-node simulated cluster on an F1 Instance and run benchmarks against it. This demo simulates a pre-built “vanilla” cluster, but demonstrates FireSim’s high performance and usability.

Why FireSim + F1?

FPGA-accelerated hardware simulation is by no means a new concept. However, previous attempts to use FPGAs for simulation have been fraught with usability, scalability, and cost issues. FireSim takes advantage of EC2 F1 and open-source hardware to address the traditional problems with FPGA-accelerated simulation:
Problem #1: FPGA-based simulations have traditionally been expensive, difficult to deploy, and difficult to reproduce.
FireSim uses public-cloud infrastructure like F1, which means no upfront cost to purchase and deploy FPGAs. Developers and researchers can distribute pre-built AMIs and AFIs, as in this public demo (more details later in this post), to make experiments easy to reproduce. FireSim also automates most of the work involved in deploying an FPGA simulation, essentially enabling one-click conversion from new RTL to deploying on an FPGA cluster.

Problem #2: FPGA-based simulations have traditionally been difficult (and expensive) to scale.
Because FireSim uses F1, users can scale out experiments by spinning up additional EC2 instances, rather than spending hundreds of thousands of dollars on large FPGA clusters.

Problem #3: Finding open hardware to simulate has traditionally been difficult. Finding open hardware that can run real software stacks is even harder.
FireSim simulates RocketChip, an open, silicon-proven, RISC-V-based processor platform, and adds peripherals like a NIC and disk device to build up a realistic system. Processors that implement RISC-V automatically support real operating systems (such as Linux) and even support applications like Apache and Memcached. We provide a custom Buildroot-based FireSim Linux distribution that runs on our simulated nodes and includes many popular developer tools.

Problem #4: Writing hardware in traditional HDLs is time-consuming.
Both FireSim and RocketChip use the Chisel HDL, which brings modern programming paradigms to hardware description languages. Chisel greatly simplifies the process of building large, highly parameterized hardware components.

How to use FireSim for hardware/software co-design

FireSim drastically improves the process of co-designing hardware and software by acting as a push-button interface for collaboration between hardware developers and systems software developers. The following diagram describes the workflows that hardware and software developers use when working with FireSim.

Figure 2. The FireSim custom hardware development workflow.

The hardware developer’s view:

  1. Write custom RTL for your accelerator, peripheral, or processor modification in a productive language like Chisel.
  2. Run a software simulation of your hardware design in standard gate-level simulation tools for early-stage debugging.
  3. Run FireSim build scripts, which automatically build your simulation, run it through the Vivado toolchain/AWS shell scripts, and publish an AFI.
  4. Deploy your simulation on EC2 F1 using the generated simulation driver and AFI
  5. Run real software builds released by software developers to benchmark your hardware

The software developer’s view:

  1. Deploy the AMI/AFI generated by the hardware developer on an F1 instance to simulate a cluster of nodes (or scale out to many F1 nodes for larger simulated core-counts).
  2. Connect using SSH into the simulated nodes in the cluster and boot the Linux distribution included with FireSim. This distribution is easy to customize, and already supports many standard software packages.
  3. Directly prototype your software using the same exact interfaces that the software will see when deployed on the real future system you’re prototyping, with the same performance characteristics as observed from software, even at scale.

FireSim demo v1.0

Figure 3. Cluster topology simulated by FireSim demo v1.0.

This first public demo of FireSim focuses on the aforementioned “software-developer’s view” of the custom hardware development cycle. The demo simulates a cluster of 1 to 8 RocketChip-based nodes, interconnected by a functional network simulation. The simulated nodes work just like “real” machines:  they boot Linux, you can connect to them using SSH, and you can run real applications on top. The nodes can see each other (and the EC2 F1 instance on which they’re deployed) on the network and communicate with one another. While the demo currently simulates a pre-built “vanilla” cluster, the entire hardware configuration of these simulated nodes can be modified after FireSim is open-sourced.

In this post, I walk through bringing up a single-node FireSim simulation for experienced EC2 F1 users. For more detailed instructions for new users and instructions for running a larger 8-node simulation, see FireSim Demo v1.0 on Amazon EC2 F1. Both demos walk you through setting up an instance from a demo AMI/AFI and booting Linux on the simulated nodes. The full demo instructions also walk you through an example workload, running Memcached on the simulated nodes, with YCSB as a load generator to demonstrate network functionality.

Deploying the demo on F1

In this release, we provide pre-built binaries for driving simulation from the host and a pre-built AFI that contains the FPGA infrastructure necessary to simulate a RocketChip-based node.

Starting your F1 instances

First, launch an instance using the free FireSim Demo v1.0 product available on the AWS Marketplace on an f1.2xlarge instance. After your instance has booted, log in using the user name centos. On the first login, you should see the message “FireSim network config completed.” This sets up the necessary tap interfaces and bridge on the EC2 instance to enable communicating with the simulated nodes.

AMI contents

The AMI contains a variety of tools to help you run simulations and build software for RISC-V systems, including the riscv64 toolchain, a Buildroot-based Linux distribution that runs on the simulated nodes, and the simulation driver program. For more details, see the AMI Contents section on the FireSim website.

Single-node demo

First, you need to flash the FPGA with the FireSim AFI. To do so, run:

[[email protected]_ADDR ~]$ sudo fpga-load-local-image -S 0 -I agfi-00a74c2d615134b21

To start a simulation, run the following at the command line:

[[email protected]_ADDR ~]$ boot-firesim-singlenode

This automatically calls the simulation driver, telling it to load the Linux kernel image and root filesystem for the Linux distro. This produces output similar to the following:

Simulations Started. You can use the UART console of each simulated node by attaching to the following screens:

There is a screen on:

2492.fsim0      (Detached)

1 Socket in /var/run/screen/S-centos.

You could connect to the simulated UART console by connecting to this screen, but instead opt to use SSH to access the node instead.

First, ping the node to make sure it has come online. This is currently required because nodes may get stuck at Linux boot if the NIC does not receive any network traffic. For more information, see Troubleshooting/Errata. The node is always assigned the IP address 192.168.1.10:

[[email protected]_ADDR ~]$ ping 192.168.1.10

This should eventually produce the following output:

PING 192.168.1.10 (192.168.1.10) 56(84) bytes of data.

From 192.168.1.1 icmp_seq=1 Destination Host Unreachable

64 bytes from 192.168.1.10: icmp_seq=1 ttl=64 time=2017 ms

64 bytes from 192.168.1.10: icmp_seq=2 ttl=64 time=1018 ms

64 bytes from 192.168.1.10: icmp_seq=3 ttl=64 time=19.0 ms

At this point, you know that the simulated node is online. You can connect to it using SSH with the user name root and password firesim. It is also convenient to make sure that your TERM variable is set correctly. In this case, the simulation expects TERM=linux, so provide that:

[[email protected]_ADDR ~]$ TERM=linux ssh [email protected]

The authenticity of host ‘192.168.1.10 (192.168.1.10)’ can’t be established.

ECDSA key fingerprint is 63:e9:66:d0:5c:06:2c:1d:5c:95:33:c8:36:92:30:49.

Are you sure you want to continue connecting (yes/no)? yes

Warning: Permanently added ‘192.168.1.10’ (ECDSA) to the list of known hosts.

[email protected]’s password:

#

At this point, you’re connected to the simulated node. Run uname -a as an example. You should see the following output, indicating that you’re connected to a RISC-V system:

# uname -a

Linux buildroot 4.12.0-rc2 #1 Fri Aug 4 03:44:55 UTC 2017 riscv64 GNU/Linux

Now you can run programs on the simulated node, as you would with a real machine. For an example workload (running YCSB against Memcached on the simulated node) or to run a larger 8-node simulation, see the full FireSim Demo v1.0 on Amazon EC2 F1 demo instructions.

Finally, when you are finished, you can shut down the simulated node by running the following command from within the simulated node:

# poweroff

You can confirm that the simulation has ended by running screen -ls, which should now report that there are no detached screens.

Future plans

At Berkeley, we’re planning to keep improving the FireSim platform to enable our own research in future data center architectures, like FireBox. The FireSim platform will eventually support more sophisticated processors, custom accelerators (such as Hwacha), network models, and peripherals, in addition to scaling to larger numbers of FPGAs. In the future, we’ll open source the entire platform, including Midas, the tool used to transform RTL into FPGA simulators, allowing users to modify any part of the hardware/software stack. Follow @firesimproject on Twitter to stay tuned to future FireSim updates.

Acknowledgements

FireSim is the joint work of many students and faculty at Berkeley: Sagar Karandikar, Donggyu Kim, Howard Mao, David Biancolin, Jack Koenig, Jonathan Bachrach, and Krste Asanović. This work is partially funded by AWS through the RISE Lab, by the Intel Science and Technology Center for Agile HW Design, and by ASPIRE Lab sponsors and affiliates Intel, Google, HPE, Huawei, NVIDIA, and SK hynix.

Now Available – Amazon Aurora with PostgreSQL Compatibility

Post Syndicated from Jeff Barr original https://aws.amazon.com/blogs/aws/now-available-amazon-aurora-with-postgresql-compatibility/

Late last year I told you about our plans to add PostgreSQL compatibility to Amazon Aurora. We launched the private beta shortly after that announcement, and followed it up earlier this year with an open preview. We’ve received lots of great feedback during the beta and the preview and have done our best to make sure that the product meets your needs and exceeds your expectations!

Now Generally Available
I am happy to report that Amazon Aurora with PostgreSQL Compatibility is now generally available and that you can use it today in four AWS Regions, with more to follow. It is compatible with PostgreSQL 9.6.3 and scales automatically to support up to 64 TB of storage, with 6-way replication behind the scenes to improve performance and availability.

Just like Amazon Aurora with MySQL compatibility, this edition is fully managed and is very easy to set up and to use. On the performance side, you can expect up to 3x the throughput that you’d get if you ran PostgreSQL on your own (you can read Amazon Aurora: Design Considerations for High Throughput Cloud-Native Relational Databases to learn more about how we did this).

You can launch a PostgreSQL-compatible Amazon Aurora instance from the RDS Console by selecting Amazon Aurora as the engine and PostgreSQL-compatible as the edition, and clicking on Next:

Then choose your instance class, single or Multi-AZ deployment (good for dev/test and production, respectively), set the instance name, and the administrator credentials, and click on Next:

You can choose between six instance classes (2 to 64 vCPUs and 15.25 to 488 GiB of memory):

The db.r4 instance class is new addition to Aurora and to RDS, and gives you an additional size at the top-end. The db.r4.16xlarge will give you additional write performance, and may allow you to use a single Aurora database instead of two or more sharded databases.

You can also set many advanced options on the next page, starting with network options such as the VPC and public accessibility:

You can set the cluster name and other database options. Encryption is easy to use and enabled by default; you can use the built-in default master key or choose one of your own:

You can also set failover behavior, the retention period for snapshot backups, and choose to enable collection of detailed (OS-level) metrics via Enhanced Monitoring:

After you have set it up to your liking, click on Launch DB Instance to proceed!

The new instances (primary and secondary since I specified Multi-AZ) are up and running within minutes:

Each PostgreSQL-compatible instance publishes 44 metrics to CloudWatch automatically:

With enhanced monitoring enabled, each instance collects additional per-instance and per-process metrics. It can be enabled when the instance is launched, or afterward, via Modify Instance. Here are some of the metrics collected when enhanced monitoring is enabled:

Clicking on Manage Graphs lets you choose which metrics are shown:

Per-process metrics are also available:

You can scale your read capacity by creating up to 15 Aurora replicas:

The cluster provides a single reader endpoint that you can access in order to load-balance requests across the replicas:

Performance Insights
As I noted earlier, Performance Insights is turned on automatically. This Amazon Aurora feature is wired directly into the database engine and allows you to look deep inside of each query, seeing the database resources that it uses and how they contribute to the overall response time. Here’s the initial view:

I can slice the view by SQL query in order to see how many concurrent copies of each query are running:

There are more views and options than I can fit in this post; to learn more take a look at Using Performance Insights.

Migrating to Amazon Aurora with PostgreSQL Compatibility
AWS Database Migration Service and the Schema Conversion Tool are ready to help you to move data stored in commercial and open-source databases to Amazon Aurora. The Schema Conversion Tool will perform a quick assessment of your database schemas and your code in order to help you to choose between MySQL and PostgreSQL. Our new, limited-time, Free DMS program allows you to use DMS and SCT to migrate to Aurora at no cost, with access to several types of DMS Instances for up to 6 months.

If you are already using PostgreSQL, you will be happy to hear that we support a long list of extensions including PostGIS and dblink.

Available Now
You can use Amazon Aurora with PostgreSQL Compatibility today in the US East (Northern Virginia), EU (Ireland), US West (Oregon), and US East (Ohio) Regions, with others to follow as soon as possible.

Jeff;

Ben’s Raspberry Pi Twilight Zone pinball hack

Post Syndicated from Alex Bate original https://www.raspberrypi.org/blog/twilight-zone-pinball-display/

When Ben North was faced with the dilemma of his nine-year-old son wanting him to watch his pinball games while, at the same time, Ben should be doing housework, he came up with a brilliant hack. Ben decided to investigate the inner workings of his twenty-year-old Twilight Zone pinball machine to convert its score display data into a video stream he could keep an eye on while working.

Ben North Raspberry Pi Twilight Zone Pinball

Ben ended up with this. Read on to find out how…

Dad? Dad! DAD!!

Kids love sharing their achievements. That’s a given. And so, after Ben introduced his son Zach to his beloved pinball machine, Zach wanted his dad to witness his progress. However, at some point Ben had to get back to the dull reality of adulting.

My son Zach, now 9, has been steadily getting better at [playing pinball], and is keen for me to watch his games. So he and I wanted a way for me to keep an eye on how his game is going, while I do other jobs elsewhere.

The two of them thought that, with the right tools and some fiddling, they could hijack the machine’s score information on its way to the dot matrix display and divert it to a computer. “One way to do this would be to set up a webcam.” Ben explains on his blog, “But where’s the fun in that?”

Twilight Zone pinball wizardry

After researching how the dot matrix receives and displays the score data, Ben and Zach figured out how to fetch its output using a 16-channel USB logic analyser. Then they dove into learning to convert the data the logic analyser outputs back into images.

Ben North Raspberry Pi Twilight Zone Pinball

“Exploring in more detail confirmed that the data looked reasonable. We could see well-distinguished frames and rows, and within each row, the pixel data had a mixture of high (lit pixel) and low (dark pixel).”

After Ben managed to convert the signals of one frame into a human-readable pixel image, it was time to think about the hardware that could do this conversion in real time. Though he and Zach were convinced they would have to build custom hardware to complete their project, they decided to first give the Raspberry Pi a go. And it turned out that the Pi was up to the challenge!

Ben North Raspberry Pi Twilight Zone Pinball - example output

“By an amazing coincidence, the [first] frame I decoded was one showing that I am the current Lost In The Zone champion.”

To decode the first frame, Ben had written a Python script. However, he coded the program to produce a score live stream in C++, since this language is better at handling high-speed input and output. To make sure Zach would learn from the experience, Ben explained the how and why of the program to him.

I talked through with Zach what the program needed to do — detect clock edges, sample pixel data, collect rows, etc. — but then he left me to do ‘all the boring typing’.

Ben used various pieces of open-source software while working on this project, including the sigrok suite for signal analysis and the multimedia framework gstreamer for handling the live video stream to the Raspberry Pi.

Find more information about the Twilight Zone pinball build, including a lot of technical details and the code itself, on Ben’s blog.

Worthy self-promotion from Ben

“I also did an FPGA project to replicate some of the Colossus code-breaking machine used in Bletchley Park during World War II,” explained Ben in our recent emails. “with a Raspberry Pi as the host.”

Colossus computer Twilight Zone Pinball

The original Colossus, not Ben’s.
Image c/o Wikipedia

As a bit of a history nerd myself, I think this is beyond cool. And if, like me, you’d like to learn more, check out the link here.

The post Ben’s Raspberry Pi Twilight Zone pinball hack appeared first on Raspberry Pi.

How to Compete with Giants

Post Syndicated from Gleb Budman original https://www.backblaze.com/blog/how-to-compete-with-giants/

How to Compete with Giants

This post by Backblaze’s CEO and co-founder Gleb Budman is the sixth in a series about entrepreneurship. You can choose posts in the series from the list below:

  1. How Backblaze got Started: The Problem, The Solution, and the Stuff In-Between
  2. Building a Competitive Moat: Turning Challenges Into Advantages
  3. From Idea to Launch: Getting Your First Customers
  4. How to Get Your First 1,000 Customers
  5. Surviving Your First Year
  6. How to Compete with Giants

Use the Join button above to receive notification of new posts in this series.

Perhaps your business is competing in a brand new space free from established competitors. Most of us, though, start companies that compete with existing offerings from large, established companies. You need to come up with a better mousetrap — not the first mousetrap.

That’s the challenge Backblaze faced. In this post, I’d like to share some of the lessons I learned from that experience.

Backblaze vs. Giants

Competing with established companies that are orders of magnitude larger can be daunting. How can you succeed?

I’ll set the stage by offering a few sets of giants we compete with:

  • When we started Backblaze, we offered online backup in a market where companies had been offering “online backup” for at least a decade, and even the newer entrants had raised tens of millions of dollars.
  • When we built our storage servers, the alternatives were EMC, NetApp, and Dell — each of which had a market cap of over $10 billion.
  • When we introduced our cloud storage offering, B2, our direct competitors were Amazon, Google, and Microsoft. You might have heard of them.

What did we learn by competing with these giants on a bootstrapped budget? Let’s take a look.

Determine What Success Means

For a long time Apple considered Apple TV to be a hobby, not a real product worth focusing on, because it did not generate a billion in revenue. For a $10 billion per year revenue company, a new business that generates $50 million won’t move the needle and often isn’t worth putting focus on. However, for a startup, getting to $50 million in revenue can be the start of a wildly successful business.

Lesson Learned: Don’t let the giants set your success metrics.

The Advantages Startups Have

The giants have a lot of advantages: more money, people, scale, resources, access, etc. Following their playbook and attacking head-on means you’re simply outgunned. Common paths to failure are trying to build more features, enter more markets, outspend on marketing, and other similar approaches where scale and resources are the primary determinants of success.

But being a startup affords many advantages most giants would salivate over. As a nimble startup you can leverage those to succeed. Let’s breakdown nine competitive advantages we’ve used that you can too.

1. Drive Focus

It’s hard to build a $10 billion revenue business doing just one thing, and most giants have a broad portfolio of businesses, numerous products for each, and targeting a variety of customer segments in multiple markets. That adds complexity and distributes management attention.

Startups get the benefit of having everyone in the company be extremely focused, often on a singular mission, product, customer segment, and market. While our competitors sell everything from advertising to Zantac, and are investing in groceries and shipping, Backblaze has focused exclusively on cloud storage. This means all of our best people (i.e. everyone) is focused on our cloud storage business. Where is all of your focus going?

Lesson Learned: Align everyone in your company to a singular focus to dramatically out-perform larger teams.

2. Use Lack-of-Scale as an Advantage

You may have heard Paul Graham say “Do things that don’t scale.” There are a host of things you can do specifically because you don’t have the same scale as the giants. Use that as an advantage.

When we look for data center space, we have more options than our largest competitors because there are simply more spaces available with room for 100 cabinets than for 1,000 cabinets. With some searching, we can find data center space that is better/cheaper.

When a flood in Thailand destroyed factories, causing the world’s supply of hard drives to plummet and prices to triple, we started drive farming. The giants certainly couldn’t. It was a bit crazy, but it let us keep prices unchanged for our customers.

Our Chief Cloud Officer, Tim, used to work at Adobe. Because of their size, any new product needed to always launch in a multitude of languages and in global markets. Once launched, they had scale. But getting any new product launched was incredibly challenging.

Lesson Learned: Use lack-of-scale to exploit opportunities that are closed to giants.

3. Build a Better Product

This one is probably obvious. If you’re going to provide the same product, at the same price, to the same customers — why do it? Remember that better does not always mean more features. Here’s one way we built a better product that didn’t require being a bigger company.

All online backup services required customers to choose what to include in their backup. We found that this was complicated for users since they often didn’t know what needed to be backed up. We flipped the model to back up everything and allow users to exclude if they wanted to, but it was not required. This reduced the number of features/options, while making it easier and better for the user.

This didn’t require the resources of a huge company; it just required understanding customers a bit deeper and thinking about the solution differently. Building a better product is the most classic startup competitive advantage.

Lesson Learned: Dig deep with your customers to understand and deliver a better mousetrap.

4. Provide Better Service

How can you provide better service? Use your advantages. Escalations from your customer care folks to engineering can go through fewer hoops. Fixing an issue and shipping can be quicker. Access to real answers on Twitter or Facebook can be more effective.

A strategic decision we made was to have all customer support people as full-time employees in our headquarters. This ensures they are in close contact to the whole company for feedback to quickly go both ways.

Having a smaller team and fewer layers enables faster internal communication, which increases customer happiness. And the option to do things that don’t scale — such as help a customer in a unique situation — can go a long way in building customer loyalty.

Lesson Learned: Service your customers better by establishing clear internal communications.

5. Remove The Unnecessary

After determining that the industry standard EMC/NetApp/Dell storage servers would be too expensive to build our own cloud storage upon, we decided to build our own infrastructure. Many said we were crazy to compete with these multi-billion dollar companies and that it would be impossible to build a lower cost storage server. However, not only did it prove to not be impossible — it wasn’t even that hard.

One key trick? Remove the unnecessary. While EMC and others built servers to sell to other companies for a wide variety of use cases, Backblaze needed servers that only Backblaze would run, and for a single use case. As a result we could tailor the servers for our needs by removing redundancy from each server (since we would run redundant servers), and using lower-performance components (since we would get high-performance by running parallel servers).

What do your customers and use cases not need? This can trim costs and complexity while often improving the product for your use case.

Lesson Learned: Don’t think “what can we add” to what the giants offer — think “what can we remove.”

6. Be Easy

How many times have you visited a large company website, particularly one that’s not consumer-focused, only to leave saying, “Huh? I don’t understand what you do.” Keeping your website clear, and your product and pricing simple, will dramatically increase conversion and customer satisfaction. If you’re able to make it 2x easier and thus increasing your conversion by 2x, you’ve just allowed yourself to spend ½ as much acquiring a customer.

Providing unlimited data backup wasn’t specifically about providing more storage — it was about making it easier. Since users didn’t know how much data they needed to back up, charging per gigabyte meant they wouldn’t know the cost. Providing unlimited data backup meant they could just relax.

Customers love easy — and being smaller makes easy easier to deliver. Use that as an advantage in your website, marketing materials, pricing, product, and in every other customer interaction.

Lesson Learned: Ease-of-use isn’t a slogan: it’s a competitive advantage. Treat it as seriously as any other feature of your product

7. Don’t Be Afraid of Risk

Obviously unnecessary risks are unnecessary, and some risks aren’t worth taking. However, large companies that have given guidance to Wall Street with a $0.01 range on their earning-per-share are inherently going to be very risk-averse. Use risk-tolerance to open up opportunities, and adjust your tolerance level as you scale. In your first year, there are likely an infinite number of ways your business may vaporize; don’t be too worried about taking a risk that might have a 20% downside when the upside is hockey stick growth.

Using consumer-grade hard drives in our servers may have caused pain and suffering for us years down-the-line, but they were priced at approximately 50% of enterprise drives. Giants wouldn’t have considered the option. Turns out, the consumer drives performed great for us.

Lesson Learned: Use calculated risks as an advantage.

8. Be Open

The larger a company grows, the more it wants to hide information. Some of this is driven by regulatory requirements as a public company. But most of this is cultural. Sharing something might cause a problem, so let’s not. All external communication is treated as a critical press release, with rounds and rounds of editing by multiple teams and approvals. However, customers are often desperate for information. Moreover, sharing information builds trust, understanding, and advocates.

I started blogging at Backblaze before we launched. When we blogged about our Storage Pod and open-sourced the design, many thought we were crazy to share this information. But it was transformative for us, establishing Backblaze as a tech thought leader in storage and giving people a sense of how we were able to provide our service at such a low cost.

Over the years we’ve developed a culture of being open internally and externally, on our blog and with the press, and in communities such as Hacker News and Reddit. Often we’ve been asked, “why would you share that!?” — but it’s the continual openness that builds trust. And that culture of openness is incredibly challenging for the giants.

Lesson Learned: Overshare to build trust and brand where giants won’t.

9. Be Human

As companies scale, typically a smaller percent of founders and executives interact with customers. The people who build the company become more hidden, the language feels “corporate,” and customers start to feel they’re interacting with the cliche “faceless, nameless corporation.” Use your humanity to your advantage. From day one the Backblaze About page listed all the founders, and my email address. While contacting us shouldn’t be the first path for a customer support question, I wanted it to be clear that we stand behind the service we offer; if we’re doing something wrong — I want to know it.

To scale it’s important to have processes and procedures, but sometimes a situation falls outside of a well-established process. While we want our employees to follow processes, they’re still encouraged to be human and “try to do the right thing.” How to you strike this balance? Simon Sinek gives a good talk about it: make your employees feel safe. If employees feel safe they’ll be human.

If your customer is a consumer, they’ll appreciate being treated as a human. Even if your customer is a corporation, the purchasing decision-makers are still people.

Lesson Learned: Being human is the ultimate antithesis to the faceless corporation.

Build Culture to Sustain Your Advantages at Scale

Presumably the goal is not to always be competing with giants, but to one day become a giant. Does this mean you’ll lose all of these advantages? Some, yes — but not all. Some of these advantages are cultural, and if you build these into the culture from the beginning, and fight to keep them as you scale, you can keep them as you become a giant.

Tesla still comes across as human, with Elon Musk frequently interacting with people on Twitter. Apple continues to provide great service through their Genius Bar. And, worst case, if you lose these at scale, you’ll still have the other advantages of being a giant such as money, people, scale, resources, and access.

Of course, some new startup will be gunning for you with grand ambitions, so just be sure not to get complacent. 😉

The post How to Compete with Giants appeared first on Backblaze Blog | Cloud Storage & Cloud Backup.