Date: Wed, 28 Mar 2001 10:24:45 +0930 From: Greg Lehey <grog@lemis.com> To: Brian Somers <brian@Awfulhak.org> Cc: "Michael C . Wu" <keichii@peorth.iteration.net>, dillon@FreeBSD.ORG, fs@FreeBSD.ORG, hackers@FreeBSD.ORG Subject: Re: Vinum stripe size (was: tuning a VERY heavily (30.0) loaded server) Message-ID: <20010328102445.C92853@wantadilla.lemis.com> In-Reply-To: <200103280040.f2S0eSc03825@hak.lan.Awfulhak.org>; from brian@Awfulhak.org on Wed, Mar 28, 2001 at 01:40:27AM %2B0100 References: <grog@lemis.com> <200103280040.f2S0eSc03825@hak.lan.Awfulhak.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday, 28 March 2001 at 1:40:27 +0100, Brian Somers wrote: >> On Tuesday, 27 March 2001 at 9:39:36 +0100, Brian Somers wrote: >>>> On Tuesday, 20 March 2001 at 11:11:44 -0600, Michael C . Wu wrote: >>>>> [Lengthy email, bear with me please, it is quite interesting. >>>>> This box averages 30.0 load with no problems.] >>>>> >>>>> <snip> >>>>> >>>>> Average file size is about 4K. /home/bbsusers* is on a vinum >>>>> stripe'd volume with 3 Ultra160 9G 10000RPM drives on sym0 at stripe >>>>> size 256K, Greg: I know this should be a prime number, >>>> >>>> No, there's no requirement for it to be a prime number. The only >>>> problem is that with 32 MB cylinder groups and a power of two stripe >>>> size and subdisk count, you end up with all the superblocks on one >>>> subdisk, which is a performance issue. Choose the stripe size so that >>>> the superblocks are roughly evenly distributed. >>> >>> A performance issue ? Surely you've misspelt ``reliability'' ? >> >> I don't see a reliability issue here. > > I believe the only issue with having all your superblocks on one disk > is a reliability thing - if you lose the disk you lose your > superblock(s). Ah. Lose part of your file system and you lose your file system. Vinum has other ways of making up for that problem. > Surely the only performance problem would be that of locality - if > you use the partition a bit, that ``problem'' should go away though. No, under normal circumstances all metadata updates would go to the same disk, since they're just behind the superblock. Greg -- Finger grog@lemis.com for PGP public key See complete headers for address and phone numbers To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-fs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010328102445.C92853>