From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gregory Farnum Subject: Re: Fw: re-exporting nfs Date: Mon, 11 Apr 2011 10:29:34 -0700 Message-ID: <7F19EF1A73CE4484873F71CBB2CFB16F@gmail.com> References: <19ABB379BFC8469C8A53FD5B3E4BD875@PC201010310710> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from mail-iy0-f174.google.com ([209.85.210.174]:52702 "EHLO mail-iy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754131Ab1DKR3i (ORCPT ); Mon, 11 Apr 2011 13:29:38 -0400 Received: by iyb14 with SMTP id 14so5814248iyb.19 for ; Mon, 11 Apr 2011 10:29:38 -0700 (PDT) In-Reply-To: <19ABB379BFC8469C8A53FD5B3E4BD875@PC201010310710> Content-Disposition: inline Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Yuki Cc: ceph-devel@vger.kernel.org On Saturday, April 9, 2011 at 11:45 PM, Yuki wrote: Hi=EF=BC=8C > I know ceph can re-export nfs protocol,but there are something confus= e me during the test. Here is the command I used. > "dd if=3D/dev/zeof of=3D/mnt/ceph/test.bat bs=3D4M count=3D1000" > The results are as follows: > ceph:62 MB/s > nfs:26MB/s > There are 3 OSDs,1 MON and 1 MDS in the ceph cluster. And I do this t= est using the in-kernel client. > I was confused by two things during the test on nfs. > 1. The "dd" command will call mknod to create a file,but mknod will c= reates a FIFO, character special file, or block special file with the s= pecified name.(It will call lookup_open to create a file during the tes= t on ceph,) I'm not quite sure what you're asking here? We don't have anything to d= o with maintaing dd so if you're asking about that program I really can= 't help you. ;) > 2.There are sometimes no network traffic during the test on nfs.(I us= e the command "sar" to monitor the local LAN network traffic ) I'm not personally very familiar with the nfs exporting, but my guess i= s that maybe some of the writes are being buffered up on the export mac= hine and then streamed out from there when it gets a commit from the NF= S client. In general it wouldn't surprise me if NFS re-export is a little slower = than native Ceph since NFS routes everything through a single machine w= hile Ceph talks directly to a lot of the servers. You might experiment = with a few of the other dd settings (to adjust data sync, etc) and see = how that changes your results. -Greg -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html