Date: Wed, 11 Jul 2001 19:36:18 -0500 (CDT) From: Mike Silbersack <silby@silby.com> To: Leo Bicknell <bicknell@ufp.org> Cc: <freebsd-hackers@freebsd.org> Subject: Re: Network performance tuning. Message-ID: <20010711192346.F2662-100000@achilles.silby.com> In-Reply-To: <20010711195021.A89324@ussenterprise.ufp.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 11 Jul 2001, Leo Bicknell wrote: > > I'm going to bring up a topic that is sure to spark a great debate > (read: flamefest), but I think it's an important issue. I've put > my nomex on, let's see where this goes. I don't think this will start a flamefest; most of what you suggest is definitely needed. However, the main question is one of developer time. Bosko just rewrote the mbuf subsystem in -current, making memory reclamation more feasible. However, I doubt much of this will be material that can be ported back to 4.x. You seem to have hit at the crux of the problem - we need dynamically tuned socket buffers. I think that there are patches which implement that feature for netbsd, perhaps they can be ported over. If you (or anyone else with free time) would port that code, I don't think it would see barriers to inclusion. <hint hint> As for changing the default buffer sizes and mbuf to mbuf cluster ratio... that could certainly spark long debate. In general, I agree with your suggestions. However, let's cut the debate short. Since you have a bunch of fbsd servers you can check out, track the output of netstat -m over the course of a few days on them. From this data, the answer to the ratio question should become clear. Mike "Silby" Silbersack To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010711192346.F2662-100000>