The State of the Stack

Telcos, ISPs, CDNs, OVPs, and Workflows

I find this trend toward recentralization interesting, not least because CDNs were once put in place to reduce long-haul telco data charges (which were more expensive than shorter ones) by using proxies to cache or split streams nearer their audiences, lowering the overall cost. Now CDNs have added other layers of service, creeping up or down their natural location in the stack. These services are specific to the CDN, differentiating themselves more from the Layer 2/3 telco/ISP; this means that simply stating, "We have lots of servers everywhere," is no longer the key performance indicator it once was for buying decision makers (telcos and ISPs can make much the same boast). It's now all about those extra features that make them "more than an ISP" in that they offer CDN-specific services (which ISPs typically don't), and yet, at the same time, they tread a difficult line with their own customers, who tend to be about offering a host of professional services to do with content production workflows in addition to simple hosting of the origin or source content. This work is falling more and more to the online video platforms (OVPs) and the many boutique streaming media and webcast service providers. By creeping up the stack, the CDNs often compete with niche service providers that are using the CDN for lower-layer services.

Even more interesting is the number of telcos that are now internally deploying CDNs, which is ironic, given the view that CDNs were put in place to bypass telco pricing. It implies that telcos are finding it cheaper to circumnavigate their own pricing internally by deploying a CDN rather than simply managing their network better. Even more ironic is the number of telcos that are effectively buying their own bandwidth back again from the CDN; Akamai/BT and Global Crossing/Limelight come to mind here.

Ultimately, the ability to operate a CDN application, be it centralized or distributed, is now a telecommunications skill that operates outside the telco. The telcos' own historic pricing model has given rise to a strange situation where it is now cheaper for a telco to buy its own bandwidth, bundled with a CDN/OVP session/presentation/applications layer service offering from a CDN, and sell that (often "white labeled") to their clients as its own CDN service than it is for that same telco to buy its own bandwidth internally and build its own service offering and sell that to its clients.

Equally, despite this incongruity, it demonstrates good stack strategy, with each staying close to its own layer where its real skills are. Telcos are always in a strong selling position: I foresee continued gradual growth in the media and entertainment telecoms sectors in Europe as telcos try to become IP CDNs and extend the existing portfolio of broadcast telecoms (ATM/Frame/MPLS/FacilityLine/ etc.) to include IP and streaming delivery. These they can offer to their large and established broadcast clients, which are now moving to become IP focused.

Meanwhile, the traditional CDNs will also grow, selling on the advantage that they have access to many telcos (therefore offering a higher SLA than just one telco can offer, since it always has multiple options).
So the CDNs and the telcos are creeping outside their natural layers on the stack!

Storm on the Horizon?

There is also a new category of service offering emerging quickly: clouds. In some ways, the traditional CDN could be viewed as a PaaS (platform as a service) offering platforms for content distribution and hosting in the same way that Amazon's CloudFront offers web or database hosting.

IaaS (infrastructure as a service) providers such as Amazon's EC2 give you the flexibility to completely define your own platform and pay for exactly what you use.

For example, Amazon EC2 allows me, for 60 cents/hour plus 15 cents/GB data transfer, to run a simple four-server live CDN with a distribution server located locally to East and West Coast U.S. users, European users, and, more recently, Asia-Pacific users. This simple setup potentially offers enough capacity for many thousands of viewers. By contrast, the CDN I built 10 years ago and closed at the start of this year (Global-MIX) cost many hundreds of thousands of dollars to implement and many tens of thousands of dollars a year to operate.

Is EC2 therefore a challenge to the traditional CDN? Certainly, where the elasticity of the pricing matches the elasticity of the use, clouds are more and more being considered as a way to rapidly and temporarily get "scale" for large webcasts.
On the other hand, for 24/7, steady-flow VOD traffic, better pricing deals can be obtained from CDNs without a doubt.

However, for the odd small live event, the option to use a server in EC2 saves an endless rigmarole of getting enough of your own servers online with sufficient connectivity (as a direct comparison); also, many CDNs don't find custom provisioning easy. IaaS offers complete freedom and a pay-as-you-go basis.