X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E39A053541 for ; Wed, 6 Apr 2005 18:18:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30227-05 for ; Wed, 6 Apr 2005 17:18:37 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id 7CE6B53545 for ; Wed, 6 Apr 2005 18:18:36 +0100 (BST) Received: from dyn-70-66.tor.dsl.tht.net (dyn-70-66.tor.dsl.tht.net [134.22.70.66]) by tht.net (Postfix) with ESMTP id 913D276A0C; Wed, 6 Apr 2005 13:18:40 -0400 (EDT) Subject: Re: =?ISO-8859-1?Q?R=E9f=2E?= : Re: RE : RE: Postgresql From: Rod Taylor To: bsimon@loxane.com Cc: pgsql-performance@postgresql.org In-Reply-To: References: Content-Type: text/plain Date: Wed, 06 Apr 2005 13:18:29 -0400 Message-Id: <1112807909.92363.120.camel@home> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 FreeBSD GNOME Team Port Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.009 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200504/107 X-Sequence-Number: 11504 On Wed, 2005-04-06 at 19:08 +0200, bsimon@loxane.com wrote: > > On our production server, I can insert 5000 tuples in 2100 ms. > > Single Xeon 2.6 Ghz > 2 Gigs ram > 3ware RAID 5 SATA drives array, 3 drives only :-(( > PG 8.0 - fsync off > > I do think inserting 5000 tuples in a second (i.e 5000 insert > transactions, no bulk load) can be reached with well a configured SCSI > RAID 10 array. Yeah, I think that can be done provided there is more than one worker. My limit seems to be about 1000 transactions per second each with a single insert for a single process (round trip time down the Fibre Channel is large) but running 4 simultaneously only drops throughput to about 900 per process (total of 2400 transactions per second) and the machine still seemed to have lots of oomph to spare. Also worth noting is that this test was performed on a machine which as a noise floor receives about 200 queries per second, which it was serving during the test. > Is pgcluster worth giving a try and can it be trusted for in a > production environnement ? > Will it be possible to get a sort of real-time application ? From the design of pgcluster it looks like it adds in a significant amount of additional communication so expect your throughput for a single process to drop through the floor. --