Skip site navigation (1) Skip section navigation (2)

Re: Advice on RAID card

From: Michael Stone <mstone+postgres(at)mathom(dot)us>
To: pgsql-performance(at)postgresql(dot)org
Subject: Re: Advice on RAID card
Date: 2005-09-25 18:13:52
Message-ID: 20050925181352.GA2241@mathom.us (view raw or flat)
Thread:
Lists: pgsql-performance
On Sun, Sep 25, 2005 at 01:41:06PM -0400, Greg Stark wrote:
>Also, Raid 5 is particularly inappropriate for write-heavy Database traffic.
>Raid 5 actually hurts write latency dramatically and Databases are very
>sensitive to latency.

Software raid 5 actually may have an advantage here. The main cause for
high raid5 write latency is the necessity of having blocks from each
disk available to calculate the parity. The chances of a pc with several
gigs of ram having all the blocks cached (thus not requiring any reads)
are higher than on a hardware raid with several hundred megs of ram. 

Mike Stone

In response to

pgsql-performance by date

Next:From: Ahmad FajarDate: 2005-09-25 18:14:46
Subject: Re: tsearch2 seem very slow
Previous:From: Luke LonerganDate: 2005-09-25 18:09:30
Subject: Re: Advice on RAID card

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group