X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.2]) by svr1.postgresql.org (Postfix) with ESMTP id 3CC89D1B197 for ; Mon, 10 May 2004 23:04:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (neptune.hub.org [200.46.204.2]) (amavisd-new, port 10024) with ESMTP id 65855-01 for ; Mon, 10 May 2004 23:04:32 -0300 (ADT) Received: from sccrmhc13.comcast.net (sccrmhc13.comcast.net [204.127.202.64]) by svr1.postgresql.org (Postfix) with ESMTP id 61091D1B241 for ; Mon, 10 May 2004 23:04:28 -0300 (ADT) Received: from geophile.com ([24.91.49.146]) by comcast.net (sccrmhc13) with SMTP id <2004051102043201600t0sj7e> (Authid: audrey.orenstein); Tue, 11 May 2004 02:04:32 +0000 Message-ID: <40A03435.7020300@geophile.com> Date: Mon, 10 May 2004 22:02:29 -0400 From: Jack Orenstein User-Agent: Mozilla/4.0 (compatible;MSIE 5.5; Windows 98) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Bruce Momjian Cc: pgsql-performance@postgresql.org Subject: Re: Configuring PostgreSQL to minimize impact of checkpoints References: <200405101909.i4AJ9Wv09415@candle.pha.pa.us> In-Reply-To: <200405101909.i4AJ9Wv09415@candle.pha.pa.us> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at postgresql.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200405/66 X-Sequence-Number: 6897 Bruce Momjian wrote: > jao@geophile.com wrote: > >>4. I understand that a "background writer" is being contemplated for >>7.5. Will that replace or augment the checkpoint process? Any >>comments on how that work will apply to my problem would be >>appreciated. I wouldn't mind seeing the average performance, >>(without the spikes) go up -- let's say -- 10%, in exchange for >>more uniform performance. These spikes are a real problem. > > > The background writer is designed to address your specific problem. We > will stil checkpoint, but the spike should be greatly minimized. > Thanks. Do you know when 7.5 is expected to be released? Until then, is a workaround known? Also, are the delays I'm seeing out of the ordinary? I'm looking at one case in which two successive transactions, each updating a handful of records, take 26 and 18 *seconds* (not msec) to complete. These transactions normally complete in under 30 msec. Jack Orenstein