Hi, On 2009-7-29, at 10:34, Iljitsch van Beijnum wrote:
The question is, what do we want to tell home gateway builders?
one half-baked idea is to tell them to provision buffering for some number of milliseconds of traffic (I'm using ms here instead of bytes, because that factors out link speeds) *and* use something better than FIFO (e.g., some AQM scheme) that you configure so that the target queue is a small fraction of that buffer.
But do we want to recommend RED? I seem to remember it has fallen somewhat out of favor, but is it worse than a simple tail drop queue?
I believe we want a queueing scheme that results in a short standing queue independent of how much buffer there is. Many AQM schemes (like RED) have that property, FIFO doesn't.
We also do want ECN, so that whatever AQM scheme we recommend doesn't need to induce unnecessary losses in order to drive the queue size down.
Bob Briscoe just made the argument that weighted fair queuing isolates flows from each other so they can't interact in useful ways. So recommend against it? On the other hand, WFQ would make VoIP etc work much better in the presence of some heavy flows. But not as well as in the case where we specifically prioritize the real-time traffic.
I don't think basic WFQ has the property that it builds short standing queues (but my queue fu is ten years old).
I'm guessing that we don't want to specify any default diffserv behavior...
Actually, I think recommending support for less-than-best-effort (e.g., RFC 3662) would be good.
Lars
smime.p7s
Description: S/MIME cryptographic signature
