From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264396AbUGMAIu (ORCPT ); Mon, 12 Jul 2004 20:08:50 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264401AbUGMAIu (ORCPT ); Mon, 12 Jul 2004 20:08:50 -0400 Received: from mailgate2.mysql.com ([213.136.52.47]:25993 "EHLO mailgate.mysql.com") by vger.kernel.org with ESMTP id S264396AbUGMAIp (ORCPT ); Mon, 12 Jul 2004 20:08:45 -0400 Subject: DBT2-MySQL Benchmark. "hugemem" vs "smp" kernels From: Peter Zaitsev To: linux-kernel@vger.kernel.org Cc: alexeyk@mysql.com Content-Type: text/plain Organization: MySQL Message-Id: <1089677229.15336.693.camel@abyss.home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Mon, 12 Jul 2004 17:07:10 -0700 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hi, Some half a year ago I wrote about my experience of performance of "hugemem" vs "sm" kernels for 2.4.21-9.EL kernel from RH AS 3.0 The results I got previously were smp hugemem IO BOUND: 1450TPM 1250TPM CPU BOUND: 7000TPM 4500TPM Results are in Transactions per Minute. This was significant regression and it was some kind of expected. The results with newer kernel 2.4.21-15.EL are very strange however: smp hugemem IO BOUND: 1150TPM 1650TPM CPU BOUND: 6300TPM 5000TPM Well I'm pretty happy to see "hugemem" kernel improved a lot, but what happened with SMP kernel ? Why it is slow now for IO bound case and why it became slower than 2.4.21-9.EL ? I'm using 4*Pentium Xeon 2.0Ghz w HT, 4GB RAM, 3WARE RAID10, EXT3 -- Peter Zaitsev, Senior Support Engineer MySQL AB, www.mysql.com