Choosing Between Multiple Content Delivery Networks and Load Balancing (Part III) - Introduction to CDN Load Balancing
I first learned about CDN Load Balancing back in 2007, during my days with Limelight Networks. Back then, Limelight Networks sold a Load Balancing product that took the requests of end users and sent those requests to the various CDN’s that were part of the customer’s CDN delivery portfolio. This was done based on customer defined percentages: X percent would use one CDN, Y percent would use the second CDN, Z percent would use the third CDN (if applicable), and so on. It worked well, but had two drawbacks: 1) it only made sense economically for the largest of customers; and 2) there was a restriction where the customer coud not send more than 45% of their traffic to other CDN’s.
I moved from Limelight Networks to CDNetworks in 2008 and it was at this time that I had my second encounter with CDN Load Balancing, during a meeting with the CTO at CBS Digital Media. CBS Digital Media was already using two other CDN’s but was still interested in what CDNetworks had to offer. Upon further discussion, I found out that CBS had written their own CDN Load Balancer. Their load balancer would switch CDN’s based on performance or based on billing commitments. The granularity they had for switching was on a specific URL basis. I remember thinking at the time how interesting that home grown technology was and wondering if it meant anything for the CDN industry as a whole. Was this the first real sign of commodization?
Later that year I was trying to sell CDNetworks services to Fox News. I ended up losing the deal to Level3. I was told by the VP, Executive Producer at Fox News Digital that the reason I lost the deal was because Level3 offered Fox News a way to automatically load balance between their current CDN provider – Limelight Networks and Level3, as the new provider. This really gave Fox News a safeguard. They could substantially lower their cost using the Level3 CDN, but if there was a performance blip they could automatically fail back over to Limelight Networks.
It was at this point that I got very interested in CDN Load Balancers. It’s funny how a loss will do that to you.
I began to wonder what Load Balancer’s and their functionality offer a CDN user? How much value did they provide? Were they worth the effort?
Well, read my next two blog posts to see what the real answers are. I’ll focus on Cotendo, and DynDNS, and discuss in a bit more detail each companies product or service for load balancing between multiple CDN’s.