From: "=?GB2312?Q?=B3=C2=BB=B6?=" Subject: performance problem on nfs in kernel v2.4 and v2.6 Date: Wed, 14 Dec 2005 10:14:47 +0800 Message-ID: <20051214021641.0C18CFB045@gatekeeper.ncic.ac.cn> Mime-Version: 1.0 Content-Type: text/plain; charset="GB2312" Return-path: Received: from sc8-sf-mx1-b.sourceforge.net ([10.3.1.91] helo=mail.sourceforge.net) by sc8-sf-list2.sourceforge.net with esmtp (Exim 4.30) id 1EmMAf-0002Lx-Pi for nfs@lists.sourceforge.net; Tue, 13 Dec 2005 18:15:09 -0800 Received: from gatekeeper.ncic.ac.cn ([159.226.41.188] helo=ncic.ac.cn) by mail.sourceforge.net with smtp (Exim 4.44) id 1EmMAd-0004kM-3U for nfs@lists.sourceforge.net; Tue, 13 Dec 2005 18:15:09 -0800 Received: from chenhuan (trumpet.ncic.ac.cn [159.226.41.129]) by gatekeeper.ncic.ac.cn (Postfix) with SMTP id 0C18CFB045 for ; Wed, 14 Dec 2005 10:16:41 +0800 (CST) To: nfs@lists.sourceforge.net Sender: nfs-admin@lists.sourceforge.net Errors-To: nfs-admin@lists.sourceforge.net List-Unsubscribe: , List-Id: Discussion of NFS under Linux development, interoperability, and testing. List-Post: List-Help: List-Subscribe: , List-Archive: nfs: =09 =09I did some test on nfs in linux kernel v2.6 ,but there are some= problem i can't solve. =09I need your help! =09I run iozone on nfs in kernel v2.4.21 and v2.6.11.12 ,the test= result is as follows: =09v2.4.21: write performance 106MB/s (wsize,rsize=3D8192) =09v2.6.11.12: write performance 52MB/s ; (wsize,rsize=3D32768) =09the iozone test command I use is : =09=09iozone -c -e -f /mnt/nfs/file1 -s 1g -r 2m -i 0 -w =09My question is why the performance is lower when using v2.6? =09The test platform is : =09=09client: 1G memory, =09=09server: 4G memory, 146G disk. 2cpu. =09=09net: 1000Mb =09When i use the following command to chechk why v2.6 takes much= more time, I found that the fsync operations takes much more= time. the command is: =09strace -o result -T write, close, fsync ./iozone -c -e -f= /mnt/nfs/file1 -s 1g -r 2m -i 0 -w =09the result of this command is that the time of write is= 6(ms),while the time of sync before close is 6 (s). =09So what is wrong? how can i see performance improvement of nfs= in v2.6. Is there any key value I can tune ? Or the problem= relates to the removal of flushd while introducing pdflush in= client memory management ? =09Thank you! Waiting for your reply! ------------------------------------------------------------ National Research Center for Intelligent Computing Systems Insitute of Computing Technology Chinese Academy of Sciences=A1=A1=A1=A1=A1=A1 ------------------------------------------------------------ =09=09 =A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1 =A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1huanchen@ncic.ac.cn =A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A1=A12005-12-14 ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click _______________________________________________ NFS maillist - NFS@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs