From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mikulas Patocka Subject: [PATCHES]: dm lock optimization Date: Wed, 18 Apr 2012 23:03:33 -0400 (EDT) Message-ID: Reply-To: device-mapper development Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dm-devel-bounces@redhat.com Errors-To: dm-devel-bounces@redhat.com To: "Alasdair G. Kergon" , dm-devel@redhat.com List-Id: dm-devel.ids Hi I placed dm lock optimization patches here http://people.redhat.com/mpatocka/patches/kernel/dm-lock-optimization/ The patches remove cache-line bouncing from device mapper when multiple processors submit requests simultaneously. I was able to measure performance improvement, although in a very specific setup (it is unmeasurable with normal disks). The setup to measure it is: - two quad-core Opterons - use ramdisk as a block device, leave it empty - create 11 nested dm-linear mappings on it (so that each i/o passes through 11 levels of dm-linear and then goes to the ramdisk) - run fio with 8 threads, using direct-io to read 512-byte blocks (time fio --rw=randread --size=1G --bsrange=512-512 --direct=1 --filename=/dev/mapper/dm-test --name=job1 --name=job2 --name=job3 --name=job4 --name=job5 --name=job6 --name=job7 --name=job8) With this setup it can be measured that the patches help: no patch (3.3 kernel): 68.9s patch 1: 64.3s patch 1,2: 52.3s patch 1,2,3: 45.2s patch 1,2,3,4: 41.2s patch 1,2,3,4,5: 34.0s Mikulas