X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by postgresql.org (Postfix) with ESMTP id B1C2311F65EA for ; Fri, 7 Apr 2006 19:03:02 -0300 (ADT) Received: from postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18365-09 for ; Fri, 7 Apr 2006 19:02:57 -0300 (ADT) X-Greylist: from auto-whitelisted by SQLgrey- Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by postgresql.org (Postfix) with ESMTP id 7730D11F65DD for ; Fri, 7 Apr 2006 19:02:59 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.6/8.13.6) with ESMTP id k37M2tBL017109; Fri, 7 Apr 2006 18:02:55 -0400 (EDT) To: "Luke Lonergan" cc: "Gavin Hamill" , pgsql-performance@postgresql.org Subject: Re: pg 8.1.3, AIX, huge box, painfully slow. In-reply-to: References: Comments: In-reply-to "Luke Lonergan" message dated "Fri, 07 Apr 2006 14:50:31 -0700" Date: Fri, 07 Apr 2006 18:02:55 -0400 Message-ID: <17108.1144447375@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, score=0.129 required=5 tests=[AWL=0.129] X-Spam-Score: 0.129 X-Spam-Level: X-Archive-Number: 200604/161 X-Sequence-Number: 18245 "Luke Lonergan" writes: > That said, I find typical memory bandwidth for the P4 in applications is > limited at about 2GB/s. See here for more detail: > http://www.cs.virginia.edu/stream/standard/Bandwidth.html > In fact, looking at the results there, the IBM 650m2 only gets 6GB/s > on all 8 CPUs. I wouldn't be surprised if the strange L3 cache > architecture of the IBM 650 is holding it back from streaming memory > access efficiently. Given Gavin's latest report, I'm wondering how much the IBM slows down when a spinlock operation is involved. If the memory architecture isn't good about supporting serialized access to memory, that gaudy sounding bandwidth number might have little to do with PG's real-world behavior. On the other hand, we already know that Xeons suck about as badly as can be on that same measure; could the pSeries really be worse? regards, tom lane