Business Premise: Single-site transit traffic may not be capable of handling the transit load. Further, traffic traverses potentially many network devices, increasing latency and the potential of packet loss. The CDN promises improvement:

a)By spreading web objects closer to the eyeball networks latency is reduced

b)Fewer network elements are traversed so reliability is improved

c)Congestion points in the core of the Internet are avoided

d)CDNs have the expertise, deployed infrastructure, economies of scale from aggregation efficiencies.

Pro: Remove the transit provider(s) bottleneck(s). What is important is the location of the bandwidth bottleneck. As David Cheriton says “You always want the content to be on the other side of the bottleneck (e.g. on the same side as the eyeballs).” The bottleneck will affect how large the TCP window size can grow before packet loss and poor performance occurs.

So the CDN distributes the content to the mid point in the hierarchy at the IX or within the last mile network.

Modeling VSPs using CDNs

To properly set expectations, Jeffrey Papen (Peak Web Consulting) suggests we use retail prices (2007) for modeling of VSPs using a CDN:

•Model A:  1,600Mbps @ $35 /Mbps

•Model B: 16,000Mbps @ $23 /Mbps

•Model C:160,000Mbps@ $13 /Mbps

In all three CDN models, we do not need as large a server/network infrastructure since the CDN will pull the video objects once to seed the content across the CDN. The CDN then stores the videos at or near the edge, and distributes the videos to the end users from the appropriate “pod”.

References: http://www.cdnlist.com

http://cdnpricing.com