From owner-freebsd-questions Wed Nov 3 9:50:11 1999 Delivered-To: freebsd-questions@freebsd.org Received: from s8-37-26.student.washington.edu (S8-37-26.student.washington.edu [128.208.37.26]) by hub.freebsd.org (Postfix) with ESMTP id 8D14A157E4 for ; Wed, 3 Nov 1999 09:50:01 -0800 (PST) (envelope-from jcwells@u.washington.edu) Received: from localhost (jcw@localhost) by s8-37-26.student.washington.edu (8.9.3/8.9.3) with ESMTP id WAA52560; Wed, 3 Nov 1999 22:44:11 GMT (envelope-from jcwells@u.washington.edu) X-Authentication-Warning: s8-37-26.student.washington.edu: jcw owned process doing -bs Date: Wed, 3 Nov 1999 22:44:11 +0000 (GMT) From: "Jason C. Wells" X-Sender: jcw@s8-37-26.student.washington.edu Reply-To: "Jason C. Wells" To: Jonathon McKitrick Cc: freebsd-questions@FreeBSD.ORG Subject: Re: swap space... In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Wed, 3 Nov 1999, Jonathon McKitrick wrote: >So TOO much swap space can be a bad thing? I just saw a comment saying >too much space will allow the OS to swap everything to disk and reduce >performance. The canonical answer here is 2*physical memory. I have never seen analysis to back that up, but I trust the developers to steer me right. Hmmm. Good question. I would guess that too much swap isn't really bad. It will allow you to use more memory. Now we know that programs that have to swap a lot get slowed down, to be sure. But we don't run just one program and VM doesn't merely use disc space as memory. The VM system stores most recently used data in main memory while allowing least recently used data to drop out of buffers back to disc or nevernever land. So if you had beaucoup processes with a lot of them just putting along in the background in a least recently used (sic) status, when those process were awakened, they might have to read swap and be slow. Active processes would keep data in main memory. I would think it would depend a lot on your application too. ::conjecture:: If you analyze the performance of the system in a pseudomathematical way you might find that two competing activities of fast VM versus slow swapping combine to form a "local maxima" for system performance depending on your application. It would be very cool to see this sort of analysis for wcarchive. I am sure that DG must have looked at these numbers. (Assuming DG did not care about peoples tranfer rates, bad assumption) Swap and VM is probably why maxusers went up to 6500 and then back down to 5000. I got this from "Design of the Unix OS" by Bach. There is more on this in /usr/share/doc/papers by David Greenman an the VM system. I am not a kernel hacker so people might be able to poke holes in what I have said. Thank You, | http://students.washington.edu/jcwells/ Jason Wells To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message