From: | evgeny gridasov <eugrid(at)fpm(dot)kubsu(dot)ru> |
---|---|
To: | blblack(at)gmail(dot)com |
Cc: | pgsql-performance(at)postgresql(dot)org |
Subject: | Re: Performance considerations for very heavy INSERT |
Date: | 2005-09-13 15:26:38 |
Message-ID: | 20050913192638.3093c63c.eugrid@fpm.kubsu.ru |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On Mon, 12 Sep 2005 16:04:06 -0500
Brandon Black <blblack(at)gmail(dot)com> wrote:
I've seen serveral tests PostgreSQL on JFS file system, it runs faster than using ext3.
Our production server works using JFS and RAID10,
we have 250K+ transactions per day and everything is OK.
Try switching to separate RAID10 arrays and use one for xlog, and others for data.
If you are using indexes, try to put them on separate RAID10 array.
> Test hardware right now is a dual Opteron with 4G of ram, which we've barely
> gotten 1,000 clients running against. Current disk hardware in testing is
> whatever we could scrape together (4x 3-ware PCI hardware RAID controllers,
> with 8 SATA drives in a RAID10 array off of each - aggregated up in a 4-way
> stripe with linux md driver and then formatted as ext3 with an appropriate
> stride parameter and data=writeback). Production will hopefully be a 4-8-way
> Opteron, 16 or more G of RAM, and a fiberchannel hardware raid array or two
> (~ 1TB available RAID10 storage) with 15krpm disks and battery-backed write
> cache.
--
Evgeny Gridasov
Software Developer
I-Free, Russia
From | Date | Subject | |
---|---|---|---|
Next Message | Brandon Black | 2005-09-13 15:30:59 | Re: Performance considerations for very heavy INSERT traffic |
Previous Message | Merlin Moncure | 2005-09-13 15:05:00 | Re: Low performance on Windows problem |