Date: Fri, 14 Jun 2013 12:08:28 +0200 From: Luigi Rizzo <rizzo@iet.unipi.it> To: Gleb Smirnoff <glebius@freebsd.org> Cc: Ermal Lu?i <eri@freebsd.org>, freebsd-net <freebsd-net@freebsd.org> Subject: Re: [PATH] ALTQ(9) codel algorithm implementation Message-ID: <20130614100828.GA48119@onelab2.iet.unipi.it> In-Reply-To: <20130614095125.GQ12443@FreeBSD.org> References: <CAPBZQG0XDmybC5tzchFrcf3TZC7uJ0VcKSX-cVMJ%2BsMZUUqWVg@mail.gmail.com> <20130614095125.GQ12443@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Jun 14, 2013 at 01:51:25PM +0400, Gleb Smirnoff wrote: > Ermal, ... > I'm afraid we can't grow mbuf packet header with 8 bytes just to satisfy > the ALTQ codel algo, which would definitely have a limited usage among > FreeBSD users. Thus, "enqueue_time" should go into mbuf_tags(9) not into > mbuf packet header. not to take positions one way or the other, but getting and releasing a tag on every packet is going to kill performance. If i remember well, 2-3 years ago at bsdcan there was discussion (and mention of some pending work, jeffr maybe ?) on providing some leading space in the mbuf so one could put there tags (e.g. ipfw and dummynet ones) without having to allocate them. Not sure where is this. The other thing with codel is that it needs a fairly coarse timer resolution (100us..1ms) to work so one might be happy with shorter timestamps (e.g. 4 bytes) if space allows them. cheers luigi
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130614100828.GA48119>