Date: Thu, 8 Feb 2001 09:16:04 -0800 (PST) From: Matt Dillon <dillon@earth.backplane.com> To: Geoff Buckingham <geoffb@chuggalug.clues.com> Cc: Dan Phoenix <dphoenix@bravenet.com>, Julian Elischer <julian@elischer.org>, Andrew Reilly <areilly@bigpond.net.au>, Alfred Perlstein <bright@wintelcom.net>, Andre Oppermann <oppermann@monzoon.net>, Rik van Riel <riel@conectiva.com.br>, Mike Silbersack <silby@silby.com>, Poul-Henning Kamp <phk@critter.freebsd.dk>, Charles Randall <crandall@matchlogic.com>, Jos Backus <josb@cncdsl.com>, freebsd-hackers@FreeBSD.ORG Subject: Re: vinum and qmail (RE: qmail IO problems) Message-ID: <200102081716.f18HG4T07310@earth.backplane.com> References: <3A826301.8AC3771C@elischer.org> <Pine.BSO.4.21.0102080336030.3495-100000@gandalf.bravenet.com> <20010208120831.C61928@chuggalug.clues.com>
next in thread | previous in thread | raw e-mail | index | archive | help
:A word of warning on this, when striping on an even number of drives
:using a power of 2 as as the stripe size, it is very easy to concentrate
:meta data on one drive, thereby doing away with much of your performance gain.
:
:Workarounds include striping at cluster size (16 or 32MB usually) using and
:odd number of disks (for non raid 3/5), though or experimentation.
:
:There should be a number of mails on the subject in the archives of the
:scsi mailing list, look for myself or greg lehay to find the thread.
I usually use a stripe size of 1152 for precisely that reason. It's
relatively easy to test whether the metadata is being distributed
between the two drives reasonably, just look at the drive lights while
you are newfs'ing.
-Matt
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200102081716.f18HG4T07310>