From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S268562AbUHLNfz (ORCPT ); Thu, 12 Aug 2004 09:35:55 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S268566AbUHLNfz (ORCPT ); Thu, 12 Aug 2004 09:35:55 -0400 Received: from rproxy.gmail.com ([64.233.170.194]:10817 "EHLO mproxy.gmail.com") by vger.kernel.org with ESMTP id S268562AbUHLNfg (ORCPT ); Thu, 12 Aug 2004 09:35:36 -0400 Message-ID: Date: Thu, 12 Aug 2004 23:35:35 +1000 From: Omar Kilani To: Maciej Soltysiak Subject: Re: Performance Degradation: 2.6.8-rc4-bk1 vs RHEL 2.4.21-15.0.3 Cc: linux-kernel@vger.kernel.org In-Reply-To: <295911442.20040812150922@dns.toxicfilms.tv> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <295911442.20040812150922@dns.toxicfilms.tv> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Maciej, On Thu, 12 Aug 2004 15:09:22 +0200, Maciej Soltysiak wrote: > Hi, > > Just a wild guess, please try the same tests with 2.6 kernels with > echo 0 > /proc/sys/net/ipv4/tcp_window_scaling This halves all my performance values. I've just tried 2.6.8-rc4-mm1 with the same results. I also tested the 2.6.8-rc4-mm1 kernel on a dual P4 Xeon 3.2GHz with 2MB Cache (with HT, so 4 logical processors) using 15K RPM U320 Fujitsu SCSI drives. [root@minbar root]# hdparm -T -t /dev/sda /dev/sda: Timing buffer-cache reads: 1852 MB in 2.00 seconds = 927.07 MB/sec Timing buffered disk reads: 218 MB in 3.02 seconds = 72.29 MB/sec [root@minbar root]# ./bw_tcp 127.0.0.1 0.065536 333.85 MB/sec Ran this 5 times with no real difference in value. [root@minbar root]# ab -n 100000 -c 2 http://localhost/index.html This is ApacheBench, Version 2.0.40-dev <$Revision: 1.141 $> apache-2.0 Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/ Copyright (c) 1998-2002 The Apache Software Foundation, http://www.apache.org/ Benchmarking localhost (be patient) Completed 10000 requests Completed 20000 requests Completed 30000 requests Completed 40000 requests Completed 50000 requests Completed 60000 requests Completed 70000 requests Completed 80000 requests Completed 90000 requests Finished 100000 requests Server Software: Apache/2.0.50 Server Hostname: localhost Server Port: 80 Document Path: /index.html Document Length: 51200 bytes Concurrency Level: 2 Time taken for tests: 43.787872 seconds Complete requests: 100000 Failed requests: 0 Write errors: 0 Total transferred: 852032704 bytes HTML transferred: 825032704 bytes Requests per second: 2283.74 [#/sec] (mean) Time per request: 0.876 [ms] (mean) Time per request: 0.438 [ms] (mean, across all concurrent requests) Transfer rate: 19002.13 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.0 0 0 Processing: 0 0 0.1 0 12 Waiting: 0 0 0.0 0 1 Total: 0 0 0.1 0 12 Percentage of the requests served within a certain time (ms) 50% 0 66% 0 75% 0 80% 0 90% 0 95% 0 98% 0 99% 0 100% 12 (longest request) Ran this 5 times with no real difference in performance. So a single 2.8GHz P4 HT with a 2.4 kernel has better performance than a dual Xeon 3.2GHz HT with a 2.6 kernel. I think the SCSI drives eliminate the IDE results as a factor, so this looks like it's a net loopback performance issue. Maybe. :) > I am curious if it might be related to my problems with networking. > turning of TCP WS tcp_window_scaling helps my problems. > > Regards, > Maciej Regards, Omar Kilani