YouTube continues to fascinate people, even when nobody is able to watch it.

A revealing data nugget has turned up from the Internet department of Arbor Networks on the effect (or lack of it) of yesterday’s well-publicised  YouTube ‘outage’.

The interruption was relatively brief - perhaps an hour and a half - which turns out to correspond to a measurable drop in Google Internet traffic seen by the company across a random sample of 50 small and mid-size ISPs spread across the world.

As can be seen from a graph published on the blog of chief scientist Craig Labovitz, Google traffic on these providers started to fall just before 7am EDT (currently -4 hours GMT) and remained at a lower and stable level until 8.20am, corresponding to the YouTube ‘glitch’. In fact, the glitch was the second in Google traffic, which suggests the problem was preceded by an earlier,  shorter one on its video service.

Bear in mind that this is Google traffic, but Google is now a major sub-division of all Internet traffic as previous analyses from Arbor have shown.

Labovitz downlplays the traffic drop, but you could argue that had it happened at another time of the day its effect might have been amplified. Across most of the US, it coincided with night or breakfast time, a period of lower activity.

It looks pretty impressive to me. Internet traffic patterns (http, that is) are clearly and measurably affected when YouTube goes down.

I like to think of the Internet as a digital equivalent of the universe. If the physics analogy holds, sites such as YouTube and Google search are now its super-giant stars, bending gravity in unexpected ways. Traffic plots are vital in understanding all of this because they
give us a sense of where the Internet is heading and hint at weaknesses.