Re: Performance advice for a new low(er)-power server - Mailing list pgsql-performance

From Merlin Moncure
Subject Re: Performance advice for a new low(er)-power server
Date
Msg-id BANLkTimC-th1G03BHT2iCPiw1T-zTNt0Ag@mail.gmail.com
Whole thread Raw
In response to Performance advice for a new low(er)-power server  (Haestan <haestan@gmail.com>)
List pgsql-performance
On Thu, Jun 16, 2011 at 10:09 AM, Haestan <haestan@gmail.com> wrote:
> Hi,
>
> I am evaluating hardware for a new PostgreSQL server. For reasons
> concerning power consumption and available space it should not have
> more than 4 disks (in a 1U case), if possible. Now, I am not sure what
> disks to use and how to layout them to get the best performance.
>
> The cheaper option would be to buy 15k Seagate SAS disks with a 3ware
> 9750SA (battery backed) controller. Does it matter whether to use a
> 4-disk RAID10 or 2x 2-disk RAID1 (system+pg_xlog , pg_data) setup? Am
> I right that both would be faster than just using a single 2-disk
> RAID1 for everything?

with 4 drives I think your best bet is single volume raid 10 (ssd or
standard disk).

> A higher end option would be to use 2x 64G Intel X-25E SSD's with a
> LSI MegaRAID 9261 controller for pg_data and/or pg_xlog and 2x SAS
> disks for the rest. Unfortunately, these SSD are the only ones offered
> by our supplier and they don't use a supercapacitor, AFAIK. Therefore
> I would have to disable the write cache on the SSD's somehow and just
> use the cache on the controller only. Does anyone know if this will
> work or even uses such a setup?

I am not a big fan of vendors that do not allow hooking in your own
drives.  How well this setup works is going to depend on how well the
controller works with the SSD.  Still, as of today, it's probably
going to be the best performance you can get for four drives...the
x25-e remains the only SLC drive from a major vendor.

> Furthermore, the LSI MegaRAID 9261 offers CacheCade which uses SSD
> disks a as secondary tier of cache for the SAS disks. Would this
> feature make sense for a PostgreSQL server, performance wise?

I'm really skeptical about this feature.

> Thank you for any hints and inputs.

The SSD space is going to see a lot more options from Intel later this
year.  See: http://www.maximumpc.com/article/news/leaked_roadmap_points_upcoming_intel_ssds.
 If you have time, I'd consider waiting a month or so to see what
options become available.

merlin

pgsql-performance by date:

Previous
From: Magnus Hagander
Date:
Subject: Re: seq scan in the case of max() on the primary key column
Next
From: Jesper Krogh
Date:
Subject: Re: Performance advice for a new low(er)-power server