Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!news.ecn.uoknor.edu!paladin.american.edu!hookup!news.join.ad.jp!news.imnet.ad.jp!usenet.seri.re.kr!news.cais.net!newsfeed.internetmci.com!news1.erols.com!uunet!not-for-mail From: jgreco@brasil.moneng.mei.com (Joe Greco) Newsgroups: comp.unix.bsd.freebsd.misc,news.admin.technical,news.software.nntp Subject: Re: Poor performance with INN and FreeBSD. Date: 16 Apr 1996 23:36:25 -0400 Organization: Marquette Electronics, Inc. - Milwaukee, WI. Lines: 30 Sender: zorch@ftp.UU.NET Approved: zorch@uunet.UU.NET Distribution: world Message-ID: <4km8jt$heh@brasil.moneng.mei.com> References: <311F8C62.4BC4@pluto.njcc.com> <4go23v$8hp@fozzie.sun3.iaf.nl> <DpHI0o.2A6@unixhub.SLAC.Stanford.EDU> <4kea27$6f0@kelewan.dandelion.com> NNTP-Posting-Host: ftp.uu.net Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:17572 news.admin.technical:1406 news.software.nntp:21717 In news.admin.technical article <4kea27$6f0@kelewan.dandelion.com>, lnz@dandelion.com (Leonard N. Zubkoff) wrote: :Indeed, striping for news is a win if done properly. You do not want :concatenation. Consider a 14GB news spool composed of 7 2GB disks that's :only 85% full. With a concentated file system, a random I/O load will be :spread only across the first 6 disks since file systems will generally :optimize for locality. If we stripe those 7 disks instead, we are :guaranteed to be spreading the random I/O load across all the spindles. :The key point is to make sure the stripe width is set large enough that the :typical I/O transaction only needs to access one of the 7 drives. Bingo, but not many people seem to understand that :-) Under FreeBSD, I've been using "just 2048" as a somewhat arbitrary number to do this. It would probably be better to split on something like cylinder group boundaries. I've been thinking about redoing this partition: /dev/rccd0e: 8380352 sectors in 2046 cylinders of 1 tracks, 4096 sectors 4092.0MB in 128 cyl groups (16 c/g, 32.00MB/g, 7680 i/g) with a stripe size of (16 c/g * 1 tr * 4096 sec) = 65536.. that's essentially alternating every 32MB between disks. I don't know how ideal that is.. ... Joe ------------------------------------------------------------------------------- Joe Greco - Systems Administrator jgreco@ns.sol.net Solaria Public Access UNIX - Milwaukee, WI 414/546-7968