ITPRC NEWS - April 2001 - http://www.itprc.com/

As I write this I'm sitting on a train heading back home from the recent "CDN" conference in New York City. This event, organized by Penton events and Stardust, focused on promoting the new technology of "Content Delivery Networks or "CDN's." While the show did have some interesting content (pardon the pun), there seems to be a few matters at hand that must be addressed in order to fully understand the technology. These topics are addressed below.

What the heck is CDN?
Ask ten people at the show for their definition of "CDN" at it is likely you would have gotten ten different answers. Some folks think of CDN as the delivery of streaming video or television over the Internet or private networks, others define it as web-switching or content-switching, and yet others define it as ways to improve web site performance. The truth is that a CDN is a bit of all of the above and then some. In my opinion, a CDN is a network optimized for specific content, such as static web pages, transaction-based web sites, streaming media, or even real-time video/audio.

There are two general approaches to building content delivery networks, the overlay approach and the network approach.

In the overlay approach, application specific servers are deployed at various points in the network to handle the distribution of specific types of content (such as web graphics or streaming video). Core network infrastructure, including routers and switches, play no part in content delivery, short of providing basic connectivity or perhaps guaranteed quality of service for specific types of traffic. A good example of the overlay model is the CDN deployed by companies such as Akamai, Digital Island and Speedera, in which content is replicated to thousands of servers around the globe.  User requests for web content is redirected to the nearest CDN server in order to improve web site response time. The overlay approach simplifies management and opens new service opportunities since CDN service providers don't have to control the underlying network infrastructure. Offering new services via the overlay approach is as simple as distributing new code to CDN servers.

In the network approach, code is deployed into routers and switches that allow them to actually recognize specific application types and make forwarding decisions based on pre-defined policies. An example of the network approach is devices that redirect content requests to local caches or switch traffic coming into data centers to specific servers optimized to serve certain types of content. In some cases, the network approach will be jointly used with the overlay approach (for example, when a switch front-ending a server farm redirects an HTTP request to an Akamai server). IP Multicast is a good example of an early network based approach to optimizing the delivery of specific types of content.

What is CDN Peering?
Recently, several manufacturers of content delivery devices such as caches and media servers have banded together to create forums and associations to promote the idea of "content peering." In this model, dozens of service providers and enterprises would create content delivery networks and peer these networks via as-of-yet-to-be developed standards in order to provide global content distribution coverage. A new IETF working group is also underway to address this issue.  The driving forces behind efforts to establish content peering standards are hardware manufacturers and smaller CDN providers in an effort to take on industry behemoth Akamai

The problem with this approach is that service providers don't have a great record of peering thus far (think of efforts to implement guaranteed quality of service across service provider boundaries that have failed miserably). For CDN peering to work, service providers would have to deploy billing and management systems that would allow their networks to exchange traffic and enforce SLA's. Given the current reluctance among many service providers to peer unless absolutely necessary, the jury is still out on the future of CDN peering.

One area where CDN peeing might work is for enterprise customers who are building internal CDN's on their intranet. Typically these networks are used to distribute streaming or real-time video such as e-learning and/or corporate announcements. Enterprises would theoretically be able to use CDN peering standards to peer their own internal CDN networks with CDN service providers to make content available both internally and via the Internet to customers or mobile workers.

What are the market drivers?
During the CDN event, many companies exhibited products and services that allowed CDN's to deliver all sorts of exciting new applications, everything from collaborative tools to games to pay-per-view television. The only problem is that nobody seems to have figured out if there is a business case (i.e. "what will people pay for?") I for one can't possibly imagine paying to watch a movie on my PC. Gaming seems to have appeal, but a lot of the CDN systems are reliant upon carriers being able to deliver high-speed broadband services, with guaranteed service levels, to residential customers, something that has been difficult for many carriers to deliver. Given the current focus on P2P - path to profitability - content providers are going to have to provide a clear business case before service providers are going to make the investment to support these applications.

What does the future hold?
So far, the biggest benefit of content delivery networks has been in improving the performance of web sites. Whether or not content networks become the basis for a new wave of Internet-based services remains to be seen. What is needed is a killer application that justifies infrastructure deployment.

One such application might just well be music distribution. The current court battles over Napster appear to be winding down and it is entirely possible that the music sharing service will soon be shutting down. Assuming that SDMI (The Secure Digital Musical Initiative) can establish some sort of encryption method that stops illegal copyrighting, it is entirely likely that big music power houses such as BMG and Columbia will make their entire catalogs available on a pay-per download basis. Napster in its current incarnation is a peer-to-peer service; users download music directly from each other's computers. If the music companies themselves begin to make their catalogues available for download, they will need to utilize a content delivery system that insures that their servers aren't crushed under the load of millions of teenagers trying to download the latest Brittany Spears hit single. This prospect alone may require content delivery systems with far more capability than current content delivery networks.  Music companies may also choose to make videos, promotional clips, and other forms of content available to consumers.

Conclusions
Content delivery networking is a broad term that encompasses many different technologies, all with the common goal of improving Internet performance. Content delivery networks have successfully improved the end user experience for millions of web users. Their emergence as key infrastructure components to support next generation Internet services will be based on the development of a strong business case for the deployment of such services.

More Information:
The ITPRC has recently launched a page on Content Networking. This page contains links to additional information about CDN's and also hosts a CDN mailing list.  See www.itprc.com
..............................

Irwin Lazar is a Senior Consultant for The Burton Group.  He focuses on strategic planning and network architecture for Fortune 500 enterprises as well as large service providers.  He is the conference director for MPLScon and runs The MPLS Resource Center  www.mplsrc.com and The Information Technology Professional's Resource Center  www.mplsrc.com.

=================================================