In a recent blog post Mark Cuban predicted the end of the Internet. Basically saying that the edges of the network will dominate as that is the only place where the bandwidth rates people will grow to expect can be maintained. Expectations will outpace the ability to expand both the backbone and ISPs connection to that backbone. I disagreed, but it got me thinking.
I think the real solution here is that CDNs need to transition to ADNs, or application delivery networks.
Imagine that instead of delivering images and HTML to your distribution network, you instead give a copy of your application. Not a bunch of PHP or Perl code, but a virtual machine that runs what ever you want your application to be. Similar to a CDN today, when someone accesses your site they get redirected to use the application somewhere physically closer to their connection. All the AJAX calls go back and forth to this local server as the connections that need to be are aggregated up to your main site. Bandwidth is minimized and the application feels very fast independent of how many people are downloading the new Paris Hilton video.
Case Study: Google Think about Google's architecture. They have all these distributed data centers, which many people have identified as them building their own CDN. That's not really the case at all because they are putting their main application in these data centers. They are just pushing out data updates to them, and letting the search happen at the leaf nodes. Google is an ADN.
With technologies like Xen and VMWare the virtualization software is already in place. The bandwidth is already in demand. Anyone have a billion dollars they want to fund my idea with? <smile>
posted Jul 30, 2007 | permanent link
