Re: High cpu usage after many inserts

From: Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>
To: Jordan Tomkinson <jordan(at)moodle(dot)com>
Cc: Aidan Van Dyk <aidan(at)highrise(dot)ca>, Greg Smith <gsmith(at)gregsmith(dot)com>, pgsql-general(at)postgresql(dot)org
Subject: Re: High cpu usage after many inserts
Date: 2009-02-25 00:50:41
Message-ID: dcc563d10902241650k6851de6wdcdf1aa65465bd70@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On Tue, Feb 24, 2009 at 5:21 PM, Jordan Tomkinson <jordan(at)moodle(dot)com> wrote:
>
> On Wed, Feb 25, 2009 at 12:05 AM, Aidan Van Dyk <aidan(at)highrise(dot)ca> wrote:
>>
>> * Greg Smith <gsmith(at)gregsmith(dot)com> [090201 00:00]:
>>
>> > Shouldn't someone have ranted about RAID-5 by this point in the thread?
>>
>> What?  Sorry, I wasn't paying attention...
>>
>> You mean someone's actually still using RAID-5?
>>
>> ;-)
>
> What exactly is wrong with RAID5 and what should we have gone with?

RAID 5 is only suitable for situations where you need maximum storage
for minimum cost and the database is mostly / all read all the time.
Like large reporting databases. It's slow on writes, and it has a low
tolerance for dead drives (2 and it's all gone)

HOWEVER. RAID-10, which is theoretically MUCH better, is only better
if it's implemented right, and lot of cheap RAID controllers don't do
any better running RAID-10. Many of these can be put into JBOD mode
where you do RAID-10 in the kernel, or you can do RAID-1 on the card
(x sets) And RAID-0 in the kernel.

RAID-10 is almost always the right choice when you're buying good
controllers and fast drives and you want maximum performance. If you
REALLY need a lot of storage, and you have to use something like RAID
5 at least look at RAID 6.

In response to

Browse pgsql-general by date

  From Date Subject
Next Message John DeSoi 2009-02-25 00:52:58 Re: Function parameter
Previous Message Jordan Tomkinson 2009-02-25 00:44:59 Re: High cpu usage after many inserts