A few minutes ago, the RSS feed from Limelight's blog showed a new entry with more details on Limelight's network capacity during the Obama inauguration webcast. While both Akamai and Limelight put out stats after the event was over, they each used different metrics in their releases making for an apples to oranges comparison.
Akamai said the total number of simultaneous audio and video streams on their network across all of their customers for that day was 7.7 million. Limelight's release only gave out the total number of simultaneous streams for the webcast, which was 2.5 million. Days later, Akamai broke out for me the total number of simultaneous streams for just the Obama webcast at around 3.8 million. Today, Limelight's blog post says that at the same time Akamai was reporting 7.7 million simultaneous streams for all of their customers, Limelight saw just over 9 million simultaneous streams for all of their customers on their network.
Clearly this is turning into a "my CDN is bigger than your CDN" debate which quite frankly, is impossible to prove for either side. For starters, neither CDN says what the average bit rate was for the simultaneous streams on their networks. If one CDN had more audio streams, or lower bitrate streams, they could deliver more than the other and it would completely skew the numbers.
In addition, network capacity is difference than network performance. Lots of folks have capacity, not as many have good performance. We don't have any third party metric to show the performance difference between the two networks during the inauguration webcast.
Limelight's bog post points to a graph by Arbor Networks that shows traffic data from ten U.S. consumer ISP networks which says that Limelight saw an increase of 160% versus Akamai's 17%. Those numbers don't make any sense to me and quite frankly, it's really hard to tell exactly what they mean.
Network scale is important, but I think performance measurement is even more crucial. I guess the CDNs putting out all of these numbers are a start, but still, I think it's nearly impossible to compare any of them apples to apples for the Obama webcast based on such limited data.