From owner-freebsd-questions Thu Oct 22 08:25:32 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id IAA10880 for freebsd-questions-outgoing; Thu, 22 Oct 1998 08:25:32 -0700 (PDT) (envelope-from owner-freebsd-questions@FreeBSD.ORG) Received: from singularity.enigami.com (singularity.enigami.com [208.140.182.42]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id IAA10875 for ; Thu, 22 Oct 1998 08:25:30 -0700 (PDT) (envelope-from ckempf@singularity.enigami.com) Received: (from ckempf@localhost) by singularity.enigami.com (8.9.1/8.9.1) id LAA20468; Thu, 22 Oct 1998 11:24:15 -0400 (EDT) To: "Steve Friedrich" , "freebsd-questions@FreeBSD.ORG" , "Konrad Heuer" Subject: Re: Dual PII - Upgrade to 3.0-R? References: <199810221343.JAA03323@laker.net> X-Copyright: Copyright (C) 1998 Cory Kempf. All Rights Reserved X-PGP-Fingerprint: 191E 2FB7 E27D 76C3 8E79 4D26 2B3B B20F 2A9C 1E1A X-PGP-Keyloc: ; finger ckempf@enigami.com From: Cory Kempf Date: 22 Oct 1998 11:24:15 -0400 In-Reply-To: "Steve Friedrich"'s message of "Thu, 22 Oct 1998 09:42:05 -0400" Message-ID: Lines: 64 X-Mailer: Gnus v5.5/XEmacs 20.4 - "Emerald" Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG "Steve Friedrich" writes: >On Thu, 22 Oct 1998 14:29:35 +0200 (CEST), Konrad Heuer wrote: > >>I administer a rather important server with dual PII board that runs very >>stable with 2.2.6-R and *one* CPU. The second CPU waits for work in a >>cabinet nearby. Any recommendations to upgrade by now to 3.0-R and to >>plug in the second CPU? Or is it still a little bit risky? Risk is a hard one to assess... I am running 3.0 with SMP, and my machine seems to function quite well. So far, no problems. OTOH, I can afford to lose a day or two if I have to, to reinstall/rebuid/re??? the system. So for me, the potential downside is pretty minimal. On the up side, The second CPU seems to have increased performance by about 60% or so. I suspect that I/O is now my bottle neck. Some performance numbers for building the kernel. The timing was done via 'time twist', where twist is: make -j11 depend > /usr/tmp/depend.out make -j11 all > /usr/tmp/all.out make -j11 install > /usr/tmp/install.out The names refer to the number of CPUs and the number passed to -j in make (e.g. twist(2:10) was 2 CPUs, -j10): twist(1:1) 155.48s user 16.88s system 84% cpu 3:23.49 total twist(1:4) 157.16s user 19.83s system 97% cpu 3:01.00 total twist(1:6) 157.43s user 20.06s system 98% cpu 2:59.52 total twist(1:7) 157.30s user 20.21s system 98% cpu 3:00.22 total twist(1:8) 157.25s user 20.22s system 98% cpu 2:59.83 total twist(1:10) 157.21s user 20.50s system 98% cpu 2:59.92 total twist(2:4) 155.16s user 37.84s system 159% cpu 2:00.77 total twist(2:6) 155.22s user 39.12s system 171% cpu 1:53.10 total twist(2:8) 155.68s user 39.54s system 173% cpu 1:52.42 total twist(2:10) 156.23s user 39.16s system 174% cpu 1:51.81 total twist(2:11) 155.50s user 40.22s system 174% cpu 1:52.00 total twist(2:12) 155.33s user 40.45s system 175% cpu 1:51.79 total twist(2:13) 156.45s user 39.68s system 175% cpu 1:51.75 total Given that the second CPU in my case was under $300, the performance boost was pretty substantial. However, I can't say that my results will, in any way, be indicitive of yours. You may need to run the one program out there that really doesn't like SMP. Also, there is a real issue at the moment with ELF -- a lot of ports are not ELF-happy. xforms, for example: % make ===> xforms-0.88.1 is broken for ELF: a.out library only. As always, YMMV. +C -- Thinking of purchasing RAM from the Chip Merchant? Please read this first: Cory Kempf Macintosh / Unix Consulting & Software Development ckempf@enigami.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message