Received: from maia.hub.org (maia-5.hub.org [200.46.204.29]) by mail.postgresql.org (Postfix) with ESMTP id 6F1A31337B77 for ; Fri, 28 Jan 2011 13:42:52 -0400 (AST) Received: from mail.postgresql.org ([200.46.204.86]) by maia.hub.org (mx1.hub.org [200.46.204.29]) (amavisd-maia, port 10024) with ESMTP id 76984-07 for ; Fri, 28 Jan 2011 17:42:45 +0000 (UTC) X-Greylist: from auto-whitelisted by SQLgrey-1.7.6 Received: from EXHUB018-1.exch018.msoutlookonline.net (exhub018-1.exch018.msoutlookonline.net [64.78.17.16]) by mail.postgresql.org (Postfix) with ESMTP id 5A9571337B5C for ; Fri, 28 Jan 2011 13:42:44 -0400 (AST) Received: from EXVMBX018-1.exch018.msoutlookonline.net ([64.78.17.47]) by EXHUB018-1.exch018.msoutlookonline.net ([64.78.17.16]) with mapi; Fri, 28 Jan 2011 09:42:43 -0800 From: Scott Carey To: Alan Hodgson , "pgsql-performance@postgresql.org" Date: Fri, 28 Jan 2011 09:44:33 -0800 Subject: Re: How to best use 32 15k.7 300GB drives? Thread-Topic: [PERFORM] How to best use 32 15k.7 300GB drives? Thread-Index: Acu/EsQg6JR4PZouTvWqfK9Nb2bMuQ== Message-ID: In-Reply-To: <201101271611.18465@hal.medialogik.com> Accept-Language: en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: user-agent: Microsoft-MacOutlook/14.2.0.101115 acceptlanguage: en-US Content-Type: multipart/alternative; boundary="_000_C9683F121E3D5scottrichrelevancecom_" MIME-Version: 1.0 X-Virus-Scanned: Maia Mailguard 1.0.1 X-Spam-Status: No, hits=-1.899 tagged_above=-10 required=5 tests=BAYES_00=-1.9, HTML_MESSAGE=0.001 X-Spam-Level: X-Archive-Number: 201101/320 X-Sequence-Number: 42107 --_000_C9683F121E3D5scottrichrelevancecom_ Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable On 1/27/11 4:11 PM, "Alan Hodgson" > wrote: On January 27, 2011, Robert Schnabel > wrote: > So my questions are 1) am I'm crazy for doing this, 2) would you change > anything and 3) is it acceptable to put the xlog & wal (and perhaps tmp > filespace) on a different controller than everything else? Please keep > in mind I'm a geneticist who happens to know a little bit about > bioinformatics and not the reverse. :-) > Putting the WAL on a second controller does help, if you're write-heavy. I tried separating indexes and data once on one server and didn't really no= tice that it helped much. Managing the space was problematic. I would sugge= st putting those together on a single RAID-10 of all the 300GB drives (minu= s a spare). It will probably outperform separate arrays most of the time, a= nd be much easier to manage. If you go this route, I suggest two equally sized RAID 10's on different co= ntrollers fir index + data, with software raid-0 on top of that. RAID 10 w= ill max out a controller after 6 to 10 drives, usually. Using the OS RAID = 0 to aggregate the throughput of two controllers works great. WAL only has to be a little bit faster than your network in most cases. I'= ve never seen it be a bottleneck on large bulk loads if it is on its own co= ntroller with 120MB/sec write throughput. I suppose a bulk load from COPY = might stress it a bit more, but CPU ends up the bottleneck in postgres once= you have I/O hardware this capable. -- A hybrid Escalade is missing the point much in the same way that having a d= iet soda with your extra large pepperoni pizza is missing the point. --_000_C9683F121E3D5scottrichrelevancecom_ Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

On 1/27/11 4:11 PM, "Alan Hodg= son" <ahodgson@simkin.ca> w= rote:

On January 27, 2011, Robert Schnabel <schnabelr@missouri.edu> wrote:
> So my questions are = 1) am I'm crazy for doing this, 2) would you change
> anything and 3) is it acce= ptable to put the xlog & wal (and perhaps tmp
> filespace) on a different co= ntroller than everything else? Please keep
> in mind I'm a geneticist who happe= ns to know a little bit about
> bioinformatics and not the reverse. :-)
>
= Putting the WAL on a second controller does help, if you're write-heavy.
I tried separating indexes and data once on one server and didn't r= eally notice that it helped much. Managing the space was problematic. I wou= ld suggest putting those together on a single RAID-10 of all the 300GB driv= es (minus a spare). It will probably outperform separate arrays most of the= time, and be much easier to manage.

If you go this route, I suggest two equally sized RAID 10'= s on different controllers fir index + data, with software raid-0 on top of= that. RAID 10 will max out a controller after 6 to 10 drives, usuall= y. Using the OS RAID 0 to aggregate the throughput of two controllers= works great.

WAL only has to be a little bit fast= er than your network in most cases. I've never seen it be a bottlenec= k on large bulk loads if it is on its own controller with 120MB/sec write t= hroughput. I suppose a bulk load from COPY might stress it a bit more= , but CPU ends up the bottleneck in postgres once you have I/O hardware thi= s capable.

--
A hybrid Escalade is missing the point much in the same way that havin= g a diet soda with your extra large pepperoni pizza is missing the point.

--_000_C9683F121E3D5scottrichrelevancecom_--