[e2e] Fundamental Questions about Router Queue in High Speed IP Networks

[David P. Reed]

At 05:21 PM 8/22/2001 -0700, Neil Spring wrote:
>4. That queues in core routers should be provisioned for
>fairness instead of utilization and price.  fairness is
>a harder problem than this.

Queues in core routers should be provisioned for low latency.  If packets 
remain in them for very long at all, i.e. to compute some "QoS" function 
like fairness or even utilization, there is something wrong.

Or, equivalently, queueing is essentially the first stage of congestion - 
which is a bad thing.  You don't congest the network to create fairness.

The false assumption pervading this whole discussion is that it presumes a 
constant network capacity, rather than assuming that the goal of a network 
is to expand to meet all the demand that customers are willing to pay for, 
and that customers will not pay very much for long latency and high jitter.


- David
--------------------------------------------
WWW Page: http://www.reed.com/dpr.html