Return to BSD News archive
Xref: sserve comp.os.386bsd.misc:4265 comp.os.linux.misc:30679 Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!agate!howland.reston.ans.net!pipex!uunet!easix!wup-gate.wup.de!andreas From: andreas@wup.de (Andreas Klemm) Newsgroups: comp.os.386bsd.misc,comp.os.linux.misc Subject: Re: How fast? [was: ... slugish ...] Followup-To: comp.os.386bsd.misc,comp.os.linux.misc Date: 29 Nov 1994 12:28:14 GMT Organization: Wiechers & Partner Datentechnik GmbH, D-40789 Monheim Lines: 58 Message-ID: <3bf6ou$pm7@wup-gate.wup.de> References: <1994Nov28.194617.18912@system9.unisys.com> NNTP-Posting-Host: sunny.wup.de X-Newsreader: TIN [version 1.2 PL2] Alex Dumitru (alexd@system9.unisys.com) wrote: : Ok! After wadding through tons and tons of articles, I have yet : to see some definitive numbers. Several people suggested running : some test, but so far nothing. I offered on the freebsd mailingliste a few benchmark results under compareable machines. SS2, 486/33 Linux Slackware, 486/33 FreeBSD 1.1, 486/33 FreeBSD 2.0 ALPHA It would be a good idea to do that again with FreeBSD 2,0 RELEASE Linux XXXware (with XXX= Slack or LST) using a 1.0.X kernel and the experimental one, where SCSI driver or such might be improved, using another write optimization (note: as far as I know they need a apecial bdflush program for the scsi cluster code). : I have a 386sx/25 with 4Mb, with a 200+ MB IDE, Hercules card ... : It is used right now for loading (and reloading) FreeBSD to get some : notes down on the installation procedure. Not just the leading edge ;-) But should show performance differences in a minimum environment. What strikes me most are the 4 MB RAM. With that equipement I normally wouldn't do a benchmark, since you aren't sure if swapping or paging (depends on the amount of daemon programs that are executed when going into multiuser mode) give strabge results ... You wouldn't expect a benchmark of a SPARCstation with 4 or 8 MB in a Unix Magazine, too ;-) : I could use it to load other OS's and run some benchmarks, after : I finish the current task. : If anyone from the Linux side could tell me what version I should be : grabbing and from where it would be appreciated. (think I have Slackware : 2.0 on one of my CD's). Slackware 2.0 is surely a good staeting point. I'd strongly suggest to do all the benchmarks in single user mode.... : I was going to look at I/O (iozone), the Byte Unix Benchmarks, and compile : times for 3 arbitrary software packages. What about SSBA and bonnie ? Don't forget to use bonnie, too, as a disk benchmark running on a 50 MB file. It shows read write performance doing i/o blocked and char. Then Rewrite and Seek time. Really interesting values. Andreas /// -- Andreas Klemm - Wiechers & Partner Datentechnik GmbH - D-40789 Monheim phone: +49 2173 3964 161 - fax: +49 2173 3964 222 - e-mail: akl@wup.de