Return to BSD News archive
Newsgroups: comp.os.linux.advocacy,comp.unix.bsd.freebsd.misc,comp.unix.advocacy,comp.unix.misc Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.mel.connect.com.au!yarrina.connect.com.au!munnari.OZ.AU!news.hawaii.edu!ames!hookup!jussieu.fr!univ-lyon1.fr!in2p3.fr!oleane!tank.news.pipex.net!pipex!usenet.eel.ufl.edu!newsfeed.internetmci.com!news.sprintlink.net!news1!not-for-mail From: root@dyson.iquest.net (John S. Dyson) Subject: Re: Linux vs FreeBSD X-Nntp-Posting-Host: dyson.iquest.net Message-ID: <4a11rt$1k2@dyson.iquest.net> Sender: news@iquest.net (News Admin) Organization: John S. Dyson's home machine References: <489kuu$rbo@pelican.cs.ucla.edu> <49phf7$rmp@times.tfs.com> <49udlr$ov@dyson.iquest.net> <49vlp2$afu@galaxy.ucr.edu> Date: Tue, 5 Dec 1995 09:02:53 GMT Lines: 40 Xref: euryale.cc.adfa.oz.au comp.os.linux.advocacy:30090 comp.unix.bsd.freebsd.misc:10298 comp.unix.advocacy:12099 comp.unix.misc:20046 In article <49vlp2$afu@galaxy.ucr.edu>, Joe Sloan <js@dostoevsky.ucr.edu> wrote: >In article <49udlr$ov@dyson.iquest.net>, >John S. Dyson <root@dyson.iquest.net> wrote: > >>I seem to remember Nick from the mailing lists -- he said that he had >>problems with FreeBSD somehow... But the issue is that HE had problems, >>but it does not invalidate FreeBSD -- because lots of people do use FreeBSD >>in high-performance applications. > >I am a big Linux fan, but FreeBSD has earned my respect. >It is extremely solid, and everything works like a well-oiled machine. > >I have switched the pcnfs/samba servers from Linux to FreeBSD, and >sleep very well at night, but my main personal workstation remains >a Linux machine, because the fact is, Linux' combination of ext2fs >and the clever disk buffer cache design blows everything else away >for interactive, disk-intensive work. > When FreeBSD V2.2 stabilizes, try mounting the UFS filesystem async, you'll get both the faster sequential read/write perf of UFS, and the async meta-data updates of Linux. The buffer caching scheme is not where Linux is "faster", it appears to be that the meta-data updates are delayed. That can open up a window, which in practice does not hurt people often. Reads and writes to/from the buffer cache in Linux are fairly slow actually. On FreeBSD typically I get about 10-16Mbyte/sec reads on my 486/66 while using Linux on exactly the same machine I get about 5-8MBytes/sec. Using ext2fs with Linux on my machine I am lucky to get 1.6 Mbytes/sec while on FreeBSD I get about 2.2 Mbytes/sec sequential -- so ext2fs is not inherently faster, even though it appears to have a simpler on disk structure. I do have some pretty sophisticated delayed metadata code on one of my working trees for FreeBSD and am about to start studying some research papers on the subject. The goal for FreeBSD is to afford perhaps two levels of async writes -- one that is very safe, and gives the best "bang" without risking the filesystem, and a no-holds-barred approach like Linux. John dyson@freebsd.org