Post Syndicated from John Graham-Cumming original https://blog.cloudflare.com/covid-19-impacts-on-internet-traffic-seattle-italy-and-south-korea/
The last few weeks have seen unprecedented changes in how people live and work around the world. Over time more and more companies have given their employees the right to work from home, restricted business travel and, in some cases, outright sent their entire workforce home. In some countries, quarantines are in place keeping people restricted to their homes.
These changes in daily life are showing up as changes in patterns of Internet use around the world. In this blog post I take a look at changing patterns in northern Italy, South Korea and the Seattle area of Washington state.
To understand how Internet use is changing, it’s first helpful to start with what a normal pattern looks like. Here’s a chart of traffic from our Dallas point of presence in the middle of January 2020.
This is a pretty typical pattern. If you look carefully you can see that Internet use is down a little at the weekend and that Internet usage is diurnal: Internet use drops down during the night and then picks up again in the morning. The peaks occur at around 2100 local time and the troughs in the dead of night at around 0300. This sort of pattern repeats worldwide with the only real difference being whether a peak occurs in the early morning (at work) or evening (at home).
Now here’s Seattle in the first week of January this year. I’ve zoomed in to a single week so we see a little more of the bumpiness of traffic during the day but it’s pretty much the same story.
Now let’s zoom out to the time period January 15 to March 12. Here’s what the chart looks like for traffic coming from Cloudflare’s Seattle PoP over that period (the gaps in the chart are just missing data in the measurement tool I’m using).
Focus in on the beginning of the chart. Looks like the familiar diurnal pattern with quieter weekends. But around January 30 something changes. There’s a big spike of traffic and traffic stays elevated. The weekends aren’t so clear either. The first reported case of COVID-19 was on January 21 in the Seattle area.
Towards the end of February, the first deaths occurred in Washington state. In early March employees of Facebook, Microsoft and Amazon in the Seattle area were all confirmed to be infected. At this point, employers began encouraging or requiring their staff to work from home. If you focus on the last part of the chart and compare it with the first two things stand out: Internet usage has grown greatly and the night time troughs are less evident. People seem to be using the Internet more and for more hours.
Throughout the period there are also days with double spikes of traffic. If I zoom into the period March 5 to March 12 it’s interesting to compare with the week in January above.
Firstly, traffic is up about 40% and nighttime troughs are now above the levels seen in January during the day. The traffic is also spiky and continues through the weekend at similar levels to the week.
Next we can zoom in on traffic to residential ISPs in the Seattle area. Here’s a chart showing the first three days of this week (March 9 to March 11) compared to Monday to Wednesday a month prior in early February (February 10 to February 12).
Traffic to residential ISPs appears to be up about 5% month on month during the work day. We might have expected this to be higher given the number of local companies asking employees to work from home but many appear to be using VPNs that route all Internet traffic back through the corporate gateway.
Turning to northern Italy, and in particular northern Italy, where there has been a serious outbreak of COVID-19 leading to first a local quarantine and then a national one. Most of the traffic in northern Italy is served from our Milan point of presence.
For reference here’s what traffic looked like the first week in January.
A familiar pattern with peak traffic typically in the evening. Here’s traffic for March 5 to 12.
Traffic has grown by more than 30% with Internet usage up at all hours of the day and night. Another change that’s a little harder to see is that traffic is ramping up earlier in the morning than in early January. In early January traffic started rising rapidly at 0900 UTC and reach the daytime plateaus you see above around 1400 UTC. In March, we see the traffic jump up more rapidly at 0900 UTC and reach a first plateau before tending to jump up again.
Drilling into the types of domains that Italians are accessing we see changes in how people are using the Internet. Online chat systems are up 1.3x to 3x of normal usage. Video streaming appears to have roughly doubled. People are accessing news and information websites about 30% to 60% more and online gaming is up about 20%.
One final look at northern Italy. Here’s the period that covers the introduction of the first cordon sanitaire in communes in the north.
The big spike of traffic is the evening of Monday, February 24 when the first cordons sanitaire came into full effect.
Here’s the normal traffic pattern in Seoul, South Korea using the first week of January as an example of what traffic looked like before the outbreak of COVID-19:
And here’s March 5 to 12 for comparison:
There’s no huge change in traffic patterns other than that Internet traffic seen by Cloudflare is up about 5%.
Digging into the websites and APIs that people are accessing in South Korea shows some significant changes: traffic to websites offering anime streaming up over 2x, online chat up 1.2x to 1.8x and online gaming up about 30%.
In both northern Italy and South Korea traffic associated with fitness trackers is down, perhaps reflecting that people are unable to take part in their usual exercise, sports and fitness activities.
Cloudflare is watching carefully as Internet traffic patterns around the world alter as people alter their daily lives through home-working, cordon sanitaire, and social distancing. None of these traffic changes raise any concern for us. Cloudflare’s network is well provisioned to handle significant spikes in traffic. We have not seen, and do not anticipate, any impact on our network’s performance, reliability, or security globally.