Ramblings on the Internet in the Era of the CDN

December 15th, 2008 by · 1 Comment

I have been mulling over the recent moves by Velocix and the general trends in content delivery over the internet, and I can't help but think that we are entering a transition period in how the internet is structured.   We already have network providers looking to enter the CDN business, and although each is taking a different track they are all talking about CDN services becoming an integral part of the internet rather than an over-the-top service.

In the current era, an ISP connected to the internet by a) peering with some networks, and b) buying transit from the rest.  End of story, all else is elaboration on the tradeoffs between peering and transit and the associated soap operas.  However, it won't be long before most bits that users download come from a CDN.  For the moment, that means little to the ISP - the bits still come down their peering or transit connections.  If they are hosting Akamai servers, they just think of that as just another type of peering that saves them transit costs - but Akamai has competition now and that will change things.

For a moment, think about the Velocix model now in that context, suppose you are an ISP with a Velocix-based CDN.  Your CDN can help you in two ways:  to help manage growing traffic loads in your aggregation network, or to sell express lane access to content providers.  Now if you're Verizon or some other behemoth, it's the latter you care most about and other content usage can just eat cake.  But if you aren't a behemoth like Verizon and pay substantial transit costs, you might want the former.  But it only helps if Velocix traffic is big (not yet true), and it doesn't help at all with traffic coming from Akamai, Limelight, Level 3, CDNetworks, etc. What you want then is to have your CDN hook up to those other CDNs.  But CDNs aren't like ISPs, they don't peer with each other, they don't trade traffic.

What does that mean?  It means that while caching and delivery of content is quickly becoming an integral part of the internet itself, the caches don't talk to each other.  The way the internet currently works, an ISP can't build a single caching solution in order to handle its growing content needs.  If they implement Velocix, they will still need Akamai servers in their closet, perhaps Level 3 CDN Expresslane service, and similar services from every CDN whose traffic is big enough to need managing.  They can't pick just one, else content from other CDNs will suffer and their users will complain and leave.

What has to happen here someday, as content grows and caching becomes all important is that new relationships beyond peering and transit need to evolve.  ISPs will want to build, outsource, or manage one express lane, not a dozen, and they will want all content to use it.  That means their local CDN needs to exchange traffic with another CDN, but the technologies are proprietary and not compatible and there are no templates for the relationship even if the technology was there.  But nobody - except maybe Cogent - wants to see CDN services wrapped up in the wholesale transit/peering business model, because that model sucks eggs and everyone knows it.  Something new will therefore have to emerge.

So where do we go from here?  I wish I knew, I only know that as CDNs become more important, traditional transit and peering become less so.  The business relationships between CDNs and ISPs will likely come to dominate the economics of the internet, and right now there are no rules to the new game.  Yes, I know that was quite a long, semi-organized diatribe, but hey that's why I called it Telecom Ramblings - sometimes it just comes out that way.

Categories: Content Distribution · Internet Traffic

Join the Discussion!

1 Comment, Add Yours!

Leave a Comment

You may Log In to post a comment, or fill in the form to post anonymously.

  • Ramblings’ Jobs

    Post a Job - Just $99/30days
  • Event Calendar