About hash() in SFQ and DRR
Tue, 14 Sep 1999 15:43:27 -0700
> From: Huiwen Li <email@example.com>
> To: "'firstname.lastname@example.org'" <ns-users@mash.CS.Berkeley.EDU>
> Subject: About hash() in SFQ and DRR
> Date: Wed, 08 Sep 1999 17:30:10 PDT
> I saw a similar question by Roman Pletka about the hash function in SFQ and
> DRR. This is also why I was confused about FQ, SFQ, DRR. Why doesn't SFQ use
> the same method when it receives the packet? Just check its flowid and puts
> it in the corresponding queues? If we use hash functions, then packets from
> different flows could be in the same queue, right? But sometimes this is not
> what we want. If I am wrong, please give me some clarification on this.
> Thanks a lot.
I believe one of the points of SFQ is that multiple flows may hash
to the same place (the upside being that you can limit the number of
flow buckets, the downside being the obvious possibility of
collisions). For DRR and FQ, I think these use per-flow queing...