Re: questions about disk configurations
От | scott.marlowe |
---|---|
Тема | Re: questions about disk configurations |
Дата | |
Msg-id | Pine.LNX.4.33.0212091556590.1276-100000@css120.ihs.com обсуждение исходный текст |
Ответ на | Re: questions about disk configurations (Josh Berkus <josh@agliodbs.com>) |
Ответы |
Re: questions about disk configurations
|
Список | pgsql-performance |
On Mon, 9 Dec 2002, Josh Berkus wrote: > > Depesz, > > > i have a question about best harddisk configuration for postgresql > > performance. > > of course i know that: > > - scsi is better than ide > > - 2 disks are better than 1 > > - 3 disks are better than 2 > > > > i know that with 3 disks one should move xlog to one drive, index files > > to second and tables to third. > > that's clear. > > Er, no, it's not. In fact, for a 3-disk config, I reccommend: > > Disk 1: OS, swap, system logs > Disk 2: Data + Indexes > Disk 3: Transaction Log Actually, first I'd try one big RAID 5 and see how it runs. THEN I'd spend time mucking around with different configs if that wasn't fast enough. If you need x performance and get 10x with a RAID 5 then move on to more interesting problems. > > but: > > will making software raid on this discs provide performance increase or > > decrease? > > Hardware RAID can improve *read* performance, particilarly RAIDs 1, 01, and > 10. For writing, the best you can do is having it not inhibit performance. > The general testament is that *software* RAID does not improve things at all; > actually, the best that can be said for Linux Software RAID 1 is that it does > not harm performance much. Not in my experience. I'd estimate my test box with dual 18 Gig UW scsis runs about 1.5 to 1.8 times faster with the two drives in a RAID1 as if a single one is used. Bonnie confirms this. single drive can read about 25 Megs a second, a pair in a RAID1 reads at about 48 Megs a second. But as you pointed out in your reply, it's more important to look at how he's gonna drive the database. If it has to input hundreds of short queries a second, that's a whole different problem than a data warehouse with 500 people throwing 8 way joins at the data all day.
В списке pgsql-performance по дате отправления: