Received: from localhost (unknown [200.46.204.184]) by developer.postgresql.org (Postfix) with ESMTP id B041E2E003A for ; Mon, 14 Apr 2008 16:58:30 -0300 (ADT) Received: from developer.postgresql.org ([200.46.204.71]) by localhost (mx1.hub.org [200.46.204.184]) (amavisd-maia, port 10024) with ESMTP id 75299-02 for ; Mon, 14 Apr 2008 16:58:23 -0300 (ADT) X-Greylist: from auto-whitelisted by SQLgrey-1.7.5 Received: from oxford.xeocode.com (unknown [62.232.55.118]) by developer.postgresql.org (Postfix) with ESMTP id 874452E002F for ; Mon, 14 Apr 2008 16:58:27 -0300 (ADT) Received: from localhost ([127.0.0.1] helo=oxford.xeocode.com) by oxford.xeocode.com with esmtp (Exim 4.69) (envelope-from ) id 1JlUoo-00063o-3y; Mon, 14 Apr 2008 20:58:22 +0100 To: "Tom Lane" Cc: "Gaetano Mendola" , Subject: Re: shared_buffers performance In-Reply-To: <21220.1208201514@sss.pgh.pa.us> (Tom Lane's message of "Mon\, 14 Apr 2008 15\:31\:54 -0400") User-Agent: Gnus/5.11 (Gnus v5.11) Emacs/22.1 (gnu/linux) X-Draft-From: ("nnimap+mail01.enterprisedb.com:INBOX" 77081) References: <48032021.90704@gmail.com> <87r6d8hfgw.fsf@oxford.xeocode.com> <21220.1208201514@sss.pgh.pa.us> From: Gregory Stark Organization: EnterpriseDB Date: Mon, 14 Apr 2008 20:58:21 +0100 Message-ID: <874pa4gqea.fsf@oxford.xeocode.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Virus-Scanned: Maia Mailguard 1.0.1 X-Archive-Number: 200804/183 X-Sequence-Number: 29676 "Tom Lane" writes: > Gregory Stark writes: >> The transition domain where performance drops dramatically as the database >> starts to not fit in shared buffers but does still fit in filesystem cache. > > It looks to me like the knee comes where the DB no longer fits in > filesystem cache. That does seem to make a lot more sense. I think I misread the units of the size of the accounts table. Reading it again it seems to be in the 1.5G-2G range for the transition which with indexes and other tables might be starting to stress the filesystem cache -- though it still seems a little low. I think if I squint I can see another dropoff at the very small scaling numbers. That must be the point where the database is comparable to the shared buffers size. Except then I would expect the green and blue curves to be pushed to the right a bit rather than just havin a shallower slope. -- Gregory Stark EnterpriseDB http://www.enterprisedb.com Ask me about EnterpriseDB's On-Demand Production Tuning