A bit of a scuffle broke out over a new report from Arbor Networks concerning the state of the internet. Wired ran an article pegging Google's bandwidth costs at zero, which was a rather poor choice of words at best. Dan Rayburn does a good job ripping apart the article itself and Data Center Knowledge added its own two cents, suffice to say that the Wired article wasn't very well thought out. But they are hardly alone in their superficial look at bandwidth costs. The central fallacies here are ones that the press often gets tangled up in, i.e. that transit is always expensive while peering is always free, and that using dark fiber magically makes it even more free.
Networks peer with each other to reduce transit costs, but the process still costs money. If it didn't, every network would peer with every other network everywhere and be free right? In reality, peering only makes economic sense under the right conditions, else transit actually is more economical. I did find interesting that Arbor says Google is peering with Tier-1 ISPs. I have heard rumors of this, the argument being simply that they have the size to force such a relationship. But I have also been told otherwise by people who know a lot about the subject that this is silly, Google's infrastructure is not similar to those networks and hence peering isn't even a reasonable idea. Sounds to me like traditional relationships in the peering/paid-peering/transit/partial-transit world are getting more complicated.
Additionally, anyone that thinks running a national network off of someone else's dark fiber is cheap is ignoring the history many dozens of companies in the last bubble and the ensuing telecom nuclear winter who found out the hard way just how expensive it can be - sometimes more than once. Just because some of the expense is amortized over time doesn't mean it isn't an expense. It's easy to buy fiber and light it, but using it to make (or save) money is actually a great deal harder than it looks. One of the things we SHOULD have learned over the last 10 years is how to properly account for the expenses of running a network. Most network operators that are still around have learned this of course. But as the Wired article demonstrates (and they are only the most recent example), the press has not.
Google is succeeding at it because it learned those network operations lessons and is using the knowledge to handle its bandwidth costs via a unique and complex blend of peering and transit, of dark fiber and leased capacity, and a content distribution strategy of its own design. It's time to stop obsessing over Google's bandwidth bill as if it is some sort of monster in the closet that is about to burst out and eat them up. The Arbor Networks report makes clear just how huge Google's international infrastructure is, while Google's published financials make it very clear that they know what they are doing. 'Nuff said.
If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!Categories: Internet Traffic