From mboxrd@z Thu Jan 1 00:00:00 1970 From: Konrad Rzeszutek Subject: Re: An multipath performance issue on RHEL 5 Date: Tue, 13 Jan 2009 10:23:06 -0500 Message-ID: <20090113152306.GA19158@mars.virtualiron.com> References: <496C5ABB.8040602@redhat.com> Reply-To: device-mapper development Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <496C5ABB.8040602@redhat.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dm-devel-bounces@redhat.com Errors-To: dm-devel-bounces@redhat.com To: device-mapper development List-Id: dm-devel.ids On Tue, Jan 13, 2009 at 05:11:23PM +0800, dwu wrote: > A customer has done a test of multipath on RHEL 5, and he found that the > speed is 30-40MB/sec, but it can reach 160MB/sec when using EMC powerpath Does it reach that when do a test on individual disks in the setup with EMC powerpath? > > [root@clnode2 ~]# hdparm -t /dev/mapper/mpath0 > > /dev/mapper/mpath0: > Timing buffered disk reads: 118 MB in 3.01 seconds = 39.22 MB/sec > [root@clnode2 ~]# hdparm -t /dev/mapper/mpath5 > > /dev/mapper/mpath5: > Timing buffered disk reads: 132 MB in 3.04 seconds = 43.38 MB/sec > [root@clnode2 tmp]# hdparm -t /dev/sdm > > /dev/sdm: > Timing buffered disk reads: 112 MB in 3.04 seconds = 36.89 MB/sec > [root@clnode2 tmp]# hdparm -t /dev/sdaa > > /dev/sdaa: > Timing buffered disk reads: 108 MB in 3.02 seconds = 35.81 MB/sec > [root@clnode2 tmp]# hdparm -t /dev/sdf > > /dev/sdf: > Timing buffered disk reads: read() failed: Input/output error > [root@clnode2 tmp]# hdparm -t /dev/sdt > > /dev/sdt: > Timing buffered disk reads: read() failed: Input/output error > Since you did the test on the underlaying SCSI subsystem (that was the next thing to test) - which has no connection to multipath it eliminates the multipath layer. When you do the hdparam test on RHEL4 OS on those disks - are the numbers the same? Is your RHEL4 rig the same exact machine with the same exact fibre connection? You could have the RHEL5 using a 1GB connection while the RHEL4 might be using 2GB?