From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Graf Subject: Re: SFQ: Reordering? Date: Sun, 8 May 2005 20:33:10 +0200 Message-ID: <20050508183310.GU28419@postel.suug.ch> References: <7bca1cb5050506145344d16b1e@mail.gmail.com> <427BEAAE.409@trash.net> <427BF3C4.1030105@trash.net> <20050506230203.GI28419@postel.suug.ch> <427BFB72.7080407@trash.net> <20050507005851.GK28419@postel.suug.ch> <427C19C3.5030304@trash.net> <20050508115100.GQ28419@postel.suug.ch> <427E3847.3050709@trash.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Asim Shankar , netdev@oss.sgi.com Return-path: To: Patrick McHardy Content-Disposition: inline In-Reply-To: <427E3847.3050709@trash.net> Sender: netdev-bounce@oss.sgi.com Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org * Patrick McHardy <427E3847.3050709@trash.net> 2005-05-08 18:03 > Actually I was beginning to think you're right about having this > feature optional. Paul McKenney's paper on SFQ states multiple times > that perturbation can cause reordering if implemented the easy way, > the Debian sfq manpage mentions this as well. So it appears to be a > design-choice. Anyways, I suggest to make the decision when we know > what the costs are. > > BTW, this is the URL for the paper: > http://rdrop.com/users/paulmck/paper/sfq.2002.06.04.pdf Indeed, well I think it depends on the perturbation interval, complexity and cost of the hashing function and the average queue length so having a flag to toggle this feature is certainly not a bad idea. Assuming a considerable queue length and a perturbation level of 1-5 seconds it probably doesn't make sense to rehash. > I agree both make sense. Are you talking about run-time adjustable > or compile-time adjustable? For SNAT it would also be nice to be > able to use the original address, unfortunately this isn't possible > anymore since we now drop the conntrack reference early. We can compute the hash at ingress or just before we drop the reference, store it in tc_classid and then use it at the egress sfq. Needs some more thinking but should be doable.