Re: RAID card recommendation

From: Scott Carey <scott(at)richrelevance(dot)com>
To: Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>, Matthew Wakeling <matthew(at)flymine(dot)org>
Cc: "pgsql-performance(at)postgresql(dot)org" <pgsql-performance(at)postgresql(dot)org>
Subject: Re: RAID card recommendation
Date: 2009-12-02 01:37:38
Message-ID: C73B04E2.1A215%scott@richrelevance.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance


On 11/24/09 11:13 AM, "Scott Marlowe" <scott(dot)marlowe(at)gmail(dot)com> wrote:

>
> They get good reviews as well. Both manufacturers have their "star"
> performers, and their "utility" or work group class controllers. For
> what you're doing the areca 12xx or 3ware 95xx series should do fine.
>

-1 to 3ware's SATA solutions

3ware 95xx and 96xx had performance somewhere between PERC 5 (horrid) and
PERC 6 (mediocre) when I tested them with large SATA drives with RAID 10.
Haven't tried raid 6 or 5. Haven't tried the "SA" model that supports SAS.
When a competing card (Areca or Adaptec) gets 3x the sequential throughput
on an 8 disk RAID 10 and only catches up to be 60% the speed after heavy
tuning of readahead value, there's something wrong.
Random access throughput doesn't suffer like that however -- but its nice
when the I/O can sequential scan faser than postgres can read the tuples.

> As far as drives go we've been really happy with WD of late, they make
> large enterprise class SATA drives that don't pull a lot of power
> (green series) and fast SATA drives that pull a bit more but are
> faster (black series). We've used both and are quite happy with each.
> We use a pair of blacks to build slony read slaves and they're very
> fast, with write speeds of ~100MB/second and read speeds double that
> in linux under sw RAID-1
>

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Karl Denninger 2009-12-02 02:08:53 Re: RAID card recommendation
Previous Message Kevin Grittner 2009-12-01 23:36:39 Re: Order by (for 15 rows) adds 30 seconds to query time