 |
 |
On Apr 14, 2011, at 6:27 AM, Jay Hanke wrote: > On Thu, Apr 14, 2011 at 12:11 AM, Owen DeLong <[log in to unmask]> wrote: >> I doubt Akamai will use them evenly. Akamai will toss traffic wherever it is most advantageous >> to Akamai. >> > > Exactly. The traffic balance will follow internet BGP magic. > No, that's a very gross oversimplification. Akamai does a lot of traffic engineering. >> They will not likely saturate anyone's network (after all, that's also counter-productive for their >> own goals in most cases). > > Also consider that the cluster will output significantly more > bandwidth than is downloaded over the Internet the load shouldn't be > more than the benefit. In smaller deployments, I've commonly seen > bandwidth savings of 2:1 or 3:1. If things got really lopsided we > could modify the BGP attributes to even it out. We're not talking a > huge amount of bandwidth with a small cluster (100 or 200 Mb/s peak > down from the internet). > I doubt BGP metrics will actually have much effect on Akamai's traffic engineering decisions. I'm not saying I think Akamai will do bad things, but, the control of where the traffic goes will be almost entirely in the hands of Akamai and you are along for the ride. Owen > jay > > ######################################################################## > > To unsubscribe from the MICE-DISCUSS list, click the following link: > http://lists.iphouse.net/cgi-bin/wa?SUBED1=MICE-DISCUSS&A=1 ######################################################################## To unsubscribe from the MICE-DISCUSS list, click the following link: http://lists.iphouse.net/cgi-bin/wa?SUBED1=MICE-DISCUSS&A=1