Date: Fri, 11 Feb 2000 12:07:09 +1030 From: Greg Lehey <grog@lemis.com> To: Scott Hess <scott@avantgo.com>, Jon Rust <jpr@vcnet.com> Cc: freebsd-questions@FreeBSD.ORG Subject: Re: Vinum questions Message-ID: <20000211120709.C76521@freebie.lemis.com> In-Reply-To: <0ea901bf73fa$65c76d90$1e80000a@avantgo.com> References: <v04210123b4c8b1545f95@[209.239.239.22]> <0ea901bf73fa$65c76d90$1e80000a@avantgo.com> <v04210126b4c8ccdad900@[209.239.239.22]> <v04210123b4c8b1545f95@[209.239.239.22]> <0ea901bf73fa$65c76d90$1e80000a@avantgo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday, 10 February 2000 at 11:09:47 -0800, Scott Hess wrote: > "Jon Rust" <jpr@vcnet.com> >> My 2 choices: >> >> 1) Stripe and mirror. Gives 18G of storage, and best performance >> that still has protection against drive failure. Not the most >> efficient use of my drives, however. >> >> 2) RAID-5 all 4 of the drives in one volume. 27G of storage. Decent >> performance. But... stability of RAID-5 in vinum? > > Make sure that RAID5 still meets your performance needs when rebuilding a > drive. When we were deciding between RAID5 and RAID0+1 (on a hardware > SCSI-SCSI controller), most of the performance tests came out reasonably > close to each other - except when rebuilding a drive. There, RAID0+1 took > a relatively minor hit, but RAID5 dropped our tps to 1/4 of the normal > amount, meaning we couldn't rebuild a drive during peak usage. RAID-5 writes are only about 1/4 the speed of RAID-1 writes. If you're writing a lot, RAID-5 is a poor choice. On Thursday, 10 February 2000 at 14:13:37 -0800, Jon Rust wrote: > At 11:09 AM -0800 2/10/00, Scott Hess wrote: >> "Jon Rust" <jpr@vcnet.com> >>> My 2 choices: >>> >>> 1) Stripe and mirror. Gives 18G of storage, and best performance >>> that still has protection against drive failure. Not the most >>> efficient use of my drives, however. >>> >>> 2) RAID-5 all 4 of the drives in one volume. 27G of storage. Decent >>> performance. But... stability of RAID-5 in vinum? >> >> Make sure that RAID5 still meets your performance needs when rebuilding a >> drive. When we were deciding between RAID5 and RAID0+1 (on a hardware >> SCSI-SCSI controller), most of the performance tests came out reasonably >> close to each other - except when rebuilding a drive. There, RAID0+1 took >> a relatively minor hit, but RAID5 dropped our tps to 1/4 of the normal >> amount, meaning we couldn't rebuild a drive during peak usage. > > Thanks for your input. I'm leaning toward RAID 0+1. But... > > I built a RAID5 volume with the 4 drives as a test. newfs took a long > time, longer than I thought, so I decided to run some benchmarks. > iozone reported 2.7MBps. Seems kinda low. Looking through the > archives, I noticed Mr Lehey developed his own benchmark proggy > called rawio: > > sudo rawio -a /dev/vinum/webr5 > > Doh! I forgot to run it on the raw device. Hit ctrl-c... bad plan. > The whole machine locked up. Even ctrl-alt-del wouldn't do anything. > Is this a problem with rawio, vinum, or vinum RAID5? Doesn't seem > like an interrupt signal should cause this type of problem. Right, that was a bug which has since been fixed. I still don't have a really good feeling about RAID-5. It works fine, and it carries on working without a hitch if a drive dies, but recovery is still a little flaky, and I'm worried that people might corrupt their volumes beyond recovery with the 'start' on a new drive. There's a workaround: back up the volume, do a newfs, and restore it again, but it's not the cleanest. I hope to have RAID-5 recovery working properly Real Soon Now. Greg -- When replying to this message, please copy the original recipients. For more information, see http://www.lemis.com/questions.html 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-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000211120709.C76521>