|
CDNs: A Dead-end Road? One of the industry veterans I interviewed during my research for Strategies for Web Hosting and Managed Services wrote to suggest that I may want to re-think writing the next book, Internet Content Delivery.
His premise is this: "CDNs were invented when 60% of the time it takes to download a file came from WAN latency. Today, 15% of the time comes from WAN latency and the rest is from local loop. That means that if Akamai completely eliminated the WAN latency, it wouldn't be noticeable to the surfer! I also know most of the CDN companies are out of business or about to go out of business, hanging from a service provider model into a software model."
Here's my reply. "Your point about the shift of the bottleneck to the last mile is a good one. However, using the same logic, once the last-mile problem *is* resolved, the bottleneck shifts back, at least toward the cloud. I even addressed that in the first book's chapter on connectivity performance, showing how packet loss (in particular) is much more significant as the speed on the lowest link along the path increases. Although the deployments of DSL and cable are still going slowly, a few years from now content providers will be faced with deployment in a world where the local loop is not the bottleneck. Of course, the book won't even be on the shelves until mid-2002, so it may coincide reasonably well.
"That the current crop of CDNs are going out of business doesn't bother me. Like others, they're finding that they grew too quickly, driven by the abundance of seemingly endless capitalization and debt financing. We know that now, given perfect hindsight. Luckily, the technology and knowledge will survive for the next generation of players."
Ah, if only the wisdom that might protect us from greedy financing would survive as well. Unfortunately, greed has a short memory.
Posted Tuesday, August 14, 2001 6:34:27 AM 
|
