I’ve been asked several times over the last month by skibare, a regular reader and (infamous) Level 3 partisan, to discuss the bandwidth choices of the most bandwidth intensive of the current generation of websites. He has in mind MySpace, Facebook, Youtube, and Skype, but I will comment on the bandwidth choices of such sites in a more general fashion.
Large and quickly growing websites like these almost never buy all their bandwidth from one provider, they always put their bandwidth eggs in more than one basket. It is called multihoming, and it is necessary in order to avoid disruptions that inevitably happen on any backbone over time – fiber cuts, router misconfigurations, whatever. You can buy blended bandwidth from providers like Internap, but that is usually something smaller players do. The largest websites and applications generally do it themselves and always pick their bandwidth building blocks from the largest wholesale IP transit suppliers. They shift the blocks around in order to play each provider off the others to bring the pricing down, but the major result is that nobody gets more than half of the business and rarely that.
So yes I’m sure that MySpace, Facebook, Youtube, Skype, and most others use large Level 3 pipes to power their sites. I’m also sure they use large pipes from Cogent, AT&T, Global Crossing, Savvis, etc alongside and in various combinations. In a way, it highlights part of why IP transit sucks as a standalone product: there is a natural ceiling on market share that has nothing to do with cost structure and quality of service, but there is no floor. To put it another way, you can lose, but you can never win. There was a day when it was glamorous to be serving IP transit to big websites, but the business model never worked for anyone and the survivors of that war are survivors and not victors.
That’s why the healthier of the wholesale IP transit providers (AT&T, Level 3, Internap) have entered the CDN space. They seek to shift their business model, to offer IP transit as one component of a portfolio of services that also may include CDN, colocation, voip, etc, all directed at the Facebooks, Youtubes, MySpaces, and Skypes of the world. Hopefully it will work this time.
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: Content Distribution · Datacenter · Internet Backbones
Join Our LinkedIn Group
“To put it another way, you can lose, but you can never win.”
what a beartiful phrase and so apt in so many challenges of life – one I shall remember for more than a brief moment in time.
IP transit seems more like a loss leader these days. get them in the door with something simple to build a relationship, then sell them a giant backbone they don’t need (MySpace/Level3)
Rob,
IP peering is an impact hereand whether anybody wants to admit or not is why IP transit is in the toliet….
that and imho advances in ethernet technology is also suffocating the need for IP transit….
my guess is we are seeing a revolution from all IP to CoS/QoS enhanced ethernet as the new cost effective way to transport….not at the lH level yet by any means, but I see more demand right now for ethernet connections that raw IP transit or even IP-VPN….
So why won’t CDN, voip, et al go the same way with each line of business being reduced to a commodity? If the biggest, pure IP network does not provide a competitive advantage in IP transport, what is the competitive advantage in any other line of business?