Re: Integrated perc 5/i

From: "Scott Marlowe" <scott(dot)marlowe(at)gmail(dot)com>
To: "tyrrill_ed(at)emc(dot)com" <tyrrill_ed(at)emc(dot)com>
Cc: miki(at)epoch(dot)co(dot)il, pgsql-performance(at)postgresql(dot)org
Subject: Re: Integrated perc 5/i
Date: 2007-08-16 22:25:07
Message-ID: dcc563d10708161525k2b9c9905i328c599a142877f6@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On 8/16/07, tyrrill_ed(at)emc(dot)com <tyrrill_ed(at)emc(dot)com> wrote:
>
>
> Hi Miki,
>
> I am using a Dell 2950, and I recently switched from using RAID 5 of all six
> disks to three RAID 1 pairs with the OS on the first pair, postgres on the
> second except for pg_xlog, which I moved to the third pair. This
> configuration change increased the insert performance of my application by
> 40%. I have not tried RAID 10 so I cannot help you there. My suggestion is
> test both RAID 5 and RAID 10, and report back to us what you find.

Good to know.

Also, be aware that one some RAID controllers, you'll get better
performance if you make the mirrors on the RAID controller, then RAID
0 them in the OS / Kernel. RAID 0 is very low on overhead, so it
doesn't have much negative impact on the server anyway. We have a
re-purposed Dell 4600 workstation with a single CPU and 2 Gigs ram
with a 4 disk linux kernel software RAID-10 that's quite a bit faster
at most db work than the 2850 w/ dual CPUs, 4 gigs ram, and a perc 5
series controller w/ battery backed cache and a 4 disk RAID-5 it is
complementing.

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Decibel! 2007-08-16 22:41:00 Re: Integrated perc 5/i
Previous Message Merlin Moncure 2007-08-16 22:14:54 Re: Integrated perc 5/i