Parts of a Whole: Effect of COVID-19 on US Internet Traffic
Introduction
In our previous post, The Building Wave of Internet Traffic, we looked at the traffic patterns across Europe and the effect the COVID-19 pandemic has had. We examined traffic in Italy, Poland, and Spain, and demonstrated how we observed huge surges of traffic around the implementation of isolation protocols, which then reduced to more normal levels in the days after. Though, it's important to note this new level of normal is almost 30% above what we'd seen a month before at a global scale. All traffic data in this post is specific to Akamai's traffic and does not necessarily reflect traffic patterns for the Internet as a whole.
With this post, we're looking at the traffic in the United States, and specifically in the individual states most prominently affected by the crisis: Washington, Florida, California, and New York. We're also taking a look at Massachusetts, since it's where Akamai's headquartered.
A few general notes before we dig deeper into Akamai's data. First, weekends have ceased, at least from a traffic perspective. Where we used to see a significant drop in traffic every weekend, the current global trend is to see a minor dip in traffic Saturday and Sunday, if any exists at all. As mentioned briefly in the previous post, these plots were developed using UTC, aka, Greenwich Mean Time. This means the white lines denoting midnight actually represent 7pm EST and 4pm PST. If you've ever had a professor assign a project that involves timezone math, you understand why we chose not to to attempt shifting everything.
The third point to be aware of is the traffic spikes on March 3, March 10, March 27, and April 8, were produced by a European sports event stream (March 3), and downloads related to various gaming patches. While the long term effect of people working from home has driven up the traffic seen by Akamai by 30% overall, spikes like this are still responsible for a larger impact on our traffic. This is why we're seeing organizations, like Sony and Microsoft, push patches to off hours, and why streaming media providers have defaulted to lower quality streams for the time being.
The United States as a whole
While it's easy to think of the United States as a single entity, that couldn't be further from the truth, especially during the current pandemic. Any country that takes five hours to fly across is going to have regional differences. Similar to what was highlighted in our European analysis, the whole is the sum of the traffic patterns from each state. Because of this, many of the events that affect a state or region are hidden at the country level.
That said, there's still a significant value in looking at the traffic levels in the United States overall. We've seen the traffic, excluding the previously mentioned spikes, slowly build until peak on March 27, 2020 at approximately 33% above February's average traffic level. The traffic then receded to 12-15% above the average in April. Compared to the peaks of 75% over February's traffic Italy and France experienced, or the 120% spike Spain saw, this may not seem like much. But these three countries combined averaged roughly 25% as much traffic as the US did, all day, every day during the same time period.
Individual States
Similar to individual countries in the European Union, each state within the United States has been responsible for declaring a State of Emergency, or issuing Stay at Home orders on their own timeline. In some states, that responsibility has been further localized by the Governor of each state allowing the local counties and cities to make that determination. Florida is one of the most notable examples, with decisions made at the local level creating a less noticeable change in state traffic patterns.
The State of Washington was the first US state to declare a State of Emergency on February 29, and one of several states to issue a Stay at Home order on March 23, 2020. If we filter out the traffic spikes created by software downloads, the daily traffic average for Washington is nearly flat between March 10 and April 12. Weekends made little to no difference in the amount of traffic, which is a major shift by itself.
There is one way that Washington differs significantly from other states: since April 12, traffic in the state has been reduced, and is actually below what it was in the month of February. This does not appear to be linked to any specific event at the federal or state level.
Unlike Washington, the pandemic response in Florida has been conducted at both the local and state level. Individual cities and counties were allowed to develop their own plans and timing in response to COVID-19 in advance of the Governor issuing a stay at home order. This makes any strong correlation between events and traffic patterns nearly impossible. Overall, internet traffic in Florida continues to be nearly 25% greater than it was in March, which is comparable to most states.
Similar to other states, the difference between weekend traffic and weekday traffic levels has become almost nonexistent. One potential reason is that the traffic is shifting from business related activities on weekdays to streaming video and gaming activities in the evenings and weekends.
As the first to issue a state-wide Stay at Home order, California saw a subtle, but definite, increase in average daily traffic on March 19 and 20, highlighted in Figure 4. The traffic in California has often been 30% above February's average, but as April has progressed, the daily average has moved closer to 20% above that baseline. California's traffic patterns have been stable and do not reflect the decrease in traffic Akamai has seen in the state of Washington.
While New York City is the most populous center in the state (and indeed, the entire country), it is not the sum of the state of New York. Because traffic can often be served from nearby locations outside city limits, any attempt to differentiate between the city and state traffic is difficult and would be open to interpretation. We've made a conscious decision to exclude it from the scope of our analysis.
As of the time this post is being written, New York City is still the epicenter of the COVID-19 pandemic in the United States, and one of the hardest hit areas in the world. After hitting nearly 35% above the February averages on March 17 (discounting software download spikes), the state of New York has consistently experienced traffic approximately 20% above the baseline.
As home to Akamai's headquarters, the Commonwealth of Massachusetts is of special interest to us as an organization. Our internal guidance to work from home started the second week of March, shortly after a State of Emergency was declared, and only essential personnel have visited our offices since. The transition to a work from home environment for the overwhelming majority of workers has not been without hiccups, but we are fortunate enough to be an organization that was able to make such transitions smoothly. Discounting software download dates, Massachusetts has stabilized at around a 25% increase over February's average.
Conclusion
While we can state definitively that COVID-19 has driven traffic in the United States to higher levels than were expected at the beginning of 2020, the ties between local events and traffic are hard to determine with certainty. Our analysis is not meant to be a hard connection, but rather the impression that we've gathered by looking at these plots and other traffic patterns within Akamai. Because a patch download or a video stream comes from one source or has a common factor, we can make solid statements about the traffic. But that same connection to global news events is tenuous at best, and not meant to be more than a general indicator.
Internet traffic patterns across the United States are likely to persist, and if the past is a prelude to the future, we're unlikely to ever see a return to previous levels. But that's consistent with internet growth patterns for as long as Akamai has been tracking them. Traffic levels will always increase, though probably not at the explosive rate caused by the COVID-19 pandemic.
Globally, Akamai's traffic jumped by 30% in March and has stayed steady at that level. Managing traffic at high volumes is one of the areas where Akamai excels as an organization. Akamai's network is designed to work with carriers and redirect traffic from areas of congestion in order to better balance the internet.