CDN's and owning the network

May 24th, 2008 by · 5 Comments

With the entry of Level 3 and AT&T into the CDN space, a discussion of cost structure has arisen. Pure CDN's say that owning the network is not necessary and too expensive. The network operators say that owning the pipes amounts to a huge advantage. Who is right?

Well, as usual, the answer is 'it depends'. To be more specific, it depends on the structure of the CDN you are building. Akamai's CDN is huge on the edge, with tens of thousands of servers sitting in ISPs all over the place peering directly with them rather than buying IP transit. Their costs are therefore dominated by the the costs of operating servers, not by the cost of what little bandwidth they buy. Owning the backbone wouldn't save them anything material because they have designed their CDN not to use it. If Level 3 or AT&T were to reproduce Akamai's CDN, there would in fact be little advantage to owning the network.

But therein lies the actual point of disagreement. AT&T and Level 3 apparently have no intention of reproducing the Akamai system. Because they own the pipes, they have very different CDN designs which take advantage of those pipes in ways pure CDNs can't. They depend more on central nodes, and place the edge of their CDN at the edge of their network rather than in the ISPs. They will depend more on their Tier-1 status, in that they don't need to pay for *any* IP transit.

The argument that generally follows is that these nodes are 'further from the customer' and therefore worse in quality than someone like Akamai can offer. This is true, to an extent. But again, 'it depends'. If you are delivering a file to Boonton NJ, is there really much difference in the latency if you send it from 30 miles away in Equinix/Newark than if you send it from an ISP colocation 10 miles away in Morristown? Not all that much. Also, even when there is a difference, there are times when it matters (gaming) and times when it doesn't make a real difference (video streaming since it is buffered anyway). Level 3 is even more specific about this, they say they have no intention of going after the small file CDN market, it is the large files they care about. Why? Because the larger the files, the less relevant the latency and the more expensive that server heavy edge network of Akamai gets. The stability of the acronym CDN belies the dynamic nature of the technology underneath it, which has been and will be redefined regularly.

I feel the pure CDNs do have something to fear here, simply because Level 3 and AT&T have the *option* to do these things if they make sense, and Akamai and other pure CDNs don't. Having more tools at your disposal generally means one can build a better mousetrap. It would be too coincidental for Akamai to have built a network that is perfect both under its own constraints *and* under those of a network operator, common sense says that their network is optimized for the costs they see in the marketplace and nothing else. Change the constraints, and the best network design may change also.

Of course, one still has to design and build that mousetrap to be better, and Akamai looks very, very formidable on that front. Their margins are just plain huge, they run their business very well. It should be quite interesting to watch.

Categories: Content Distribution · Internet Backbones

Join the Discussion!

5 Comments So Far

  • Jim G says:

    Not to mention, for the tier 1 backbones, implementing a CDN and migrating high bandwidth customers from transit service to CDN can take traffic off their inter-city backbone, stretching those long-haul fibers and optical gear even further.

  • Parkite says:

    Akamai = Focus (on CDN).

    A jack of all trades (ATT, L3), is typically a master of none. It is striking when you review business history and find instance after instance of small, focused competitors out-executing large unfocused organizations.

  • cdn guy says:

    Rob – Glad to see this site surface. Enjoyed the posts.

    I want to point out though that you are way off base on how much transit Akamai buys. As much as they – and all CDN’s – would love to peer off 70 – 90% of their traffic, it just doesn’t happen. Most Tier1 carriers have heavy restrictions on inbound/outbound ratios and when a content heavy (outbound) company such as Akamai or LL has 10:1 ratios, the peerings just won’t come (Note that LL has their own backbone and has some eyeball traffic, so some of these peerings come easier for them).

    I can’t guess with high accuracy what percentage of traffic Akamai peers off (the claim more than it actually is), but if a gun were to my head I’d say they’d be very lucky to peer off 50%. When you’re doing 1 – 1.3 Tbps thats at least 500 – 700 Gbps of transit per month. Lets say their average transit blend of peer routes and full transit is $6/Mbps and that leaves you with a minimum bill of $3M per month or $36M per year. Not exactly peanuts.

    ISP’s will ALWAYS have a huge pricing advantage due to the fact that they don’t have to pay for transit, but Akamai, LL, Panther, and other carrier neutral CDN’s will always have the performance advantage of buying from multiple providers in more locations and choosing the “best” route for their customer. Even the biggest and best of Tier1 backbones have holes, limitations, and service problems.

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