On Wed, 2015-06-10 at 14:56 -0700, Dave Taht wrote: > I certainly would like you to try factoring in fq_codel's effects into > your calculations. 1024 queues is the default, > with 64k queues possible. If you assume 10k flows, you can readily > assume a pretty equal distribution across those available queues, with > 1k codels compensating madly. I tend to think most of the future > faster world is going to have multiple queues, with every 10gigE > device I am familiar with coming with 64 or more.
As it stands the draft doesn't say anything to this effect. > Secondly you appear to be thinking in terms of very abrupt changes in > demand not characteristic of a core link. (certainly characteristic of > a route flap, however, so I am not dissing your work). Things can change surprisingly rapidly when backhoe meets fiber. Regards, // Steve _______________________________________________ aqm mailing list aqm@ietf.org https://www.ietf.org/mailman/listinfo/aqm