Re: Load times on RAID0 compared to RAID5 - Mailing list pgsql-performance

From scott.marlowe
Subject Re: Load times on RAID0 compared to RAID5
Date
Msg-id Pine.LNX.4.33.0304070927000.14934-100000@css120.ihs.com
Whole thread Raw
In response to Load times on RAID0 compared to RAID5  (Howard Oblowitz <HowardO@LEWIS-STORES.com>)
List pgsql-performance
On Mon, 7 Apr 2003, Howard Oblowitz wrote:

> We have a perl program that loads data into Postgres version 7.3.1
> using deletes and inserts.
>
> We find that the load times are about 50% slower when we use a
> RAID5 disk system as compared to when we use RAID0
>
> Are there any postgres configuration parameters that we can set to improve
> the performance on RAID5?

Well, RAID0 SHOULD be about twice as fast as RAID5 for most applications,
maybe even faster for others.

Of course, RAID0 offers no redundancy, so if any single drive fails your
data disappears in a large puff of smoke.  RAID5 can survive a single
drive failure, and that doesn't come for free.

RAID1 may offer a better compromise of performance and reliability for
many apps than RAID5.  Generally RAID0 is fastest, RAID1 is fast but can't
grow to be as big as RAID5, RAID5 handles large parallel access better
than RAID1, RAID1 handles batch processing better than RAID5.

Mixing them together sometimes helps, sometimes not.  RAID1 on top of
RAID0 works pretty well but costs the most per meg stored than most plain
RAID5 or RAID1 setups.


pgsql-performance by date:

Previous
From: Josh Berkus
Date:
Subject: Re: Load times on RAID0 compared to RAID5
Next
From: Robert Treat
Date:
Subject: Re: ext3 filesystem / linux 7.3