One of the interesting things about working on a CDN is the ability to connect the traffic you see to real life events. This is especially prominent in the world of live streaming, where big events can contribute huge volumes of traffic. For example, our friends over at Fastly were able to see some of the specifics of the recent England Vs. Colombia world cup match play out in their per PoP traffic, a pattern that we also observed for that game.
But it turns out being a multi-tenant CDN means that sometimes we can also see what traffic isn’t happening during a live event. In other words, during a large sporting event, sometimes people stop what they’re doing to go watch that instead. Below is an example of such a scenario.
Here we consider the combined traffic of a few of our customers as seen at a single point of presence. Since I drew this plot with a blue line, we are going to go ahead and refer to this as blue content. We can see on this traffic graph, the events of the game play out in exactly opposite terms: when the game got exciting, this traffic decreased. After all was said and done, traffic corrected itself to the usual levels:
Dips in traffic closely correlate with the most exciting parts of the match.
Generally speaking, these kinds of cross-customer interactions can be extremely interesting and provide insight into internet usage that can be difficult to observe from many traditional measurement vantage points.