Re: SATA RAID: Promise vs. 3ware - Mailing list pgsql-performance

From Magnus Hagander
Subject Re: SATA RAID: Promise vs. 3ware
Date
Msg-id 20070320143448.GD24280@svr2.hagander.net
Whole thread Raw
In response to Re: SATA RAID: Promise vs. 3ware  ("Merlin Moncure" <mmoncure@gmail.com>)
List pgsql-performance
On Tue, Mar 20, 2007 at 10:18:45AM -0400, Merlin Moncure wrote:
> On 3/20/07, Ireneusz Pluta <ipluta@wp.pl> wrote:
> >Hello all,
> >
> >I sent a similar post to a FreeBSD group, but thought I'd might try here
> >too.
> >
> >I am completing a box for PostgreSQL server on FreeBSD. Selecting a RAID
> >controller I decided to go
> >with 3ware SE9650-16, following good opinions about 3ware controllers
> >found on FreeBSD and
> >PostgreSQL groups.
> >
> >However my dealer suggest me not to go with 3ware, and take Promise
> >SuperTrak EX16350, instead. This
> >suggestion does not have any technical background and it comes generally
> >from the fact of limited
> >availability of 16x 3ware controllers on the local market and immediate
> >availability of Promise.
> >
> >Is this technically a good idea to take Promise instead of 3ware or rather
> >I definitely should
> >insist on 3ware and wait for it?
>
>
> Promise raid controllers are famous for being software based with all
> the real work being done in the driver.  Without doing the research
> this may or may not be the case with this particular controller.
> Another issue with cheap RAID controllers is the performance may not
> be as good as software raid...in fact it may be worse.  Look for
> benchmarks on the web and be skeptical.

A Promise RAID is the only hardware RAID I've ever had eat an entire
array for me... Granted this was one of those "external array with SCSI
to the host", but it's certainly turned me away from Promise.. Probably
not related to the controller in question, just their general quality
level.

//Magnus


pgsql-performance by date:

Previous
From: "Merlin Moncure"
Date:
Subject: Re: SATA RAID: Promise vs. 3ware
Next
From: Vivek Khera
Date:
Subject: how small to split a table?