From mboxrd@z Thu Jan 1 00:00:00 1970 From: Phil Turmel Subject: Re: RAID performance - new kernel results - 5x SSD RAID5 Date: Sat, 02 Mar 2013 21:35:23 -0500 Message-ID: <5132B6EB.8000209@turmel.org> References: <51134E43.7090508@websitemanagers.com.au> <51137FB8.6060003@websitemanagers.com.au> <5113A2D6.20104@websitemanagers.com.au> <51150475.2020803@websitemanagers.com.au> <5120A84E.4020702@websitemanagers.com.au> <51222A81.9080600@hardwarefreak.com> <51250377.509@websitemanagers.com.au> <5125B8E5.5000502@hardwarefreak.com> <5125C154.3090603@websitemanagers.com.au> <51272808.7070302@hardwarefreak.com> <512A79D3.9020502@hardwarefreak.com> <5130D206.4090302@websitemanagers.com.au> <5131C338.8010402@hardwarefreak.com> <513231BB.4070903@turmel.org> <51328FDF.1050505@hardwarefreak.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <51328FDF.1050505@hardwarefreak.com> Sender: linux-raid-owner@vger.kernel.org To: stan@hardwarefreak.com Cc: Adam Goryachev , Linux RAID List-Id: linux-raid.ids On 03/02/2013 06:48 PM, Stan Hoeppner wrote: > On 3/2/2013 11:07 AM, Phil Turmel wrote: >> On 03/02/2013 04:15 AM, Stan Hoeppner wrote: >>> On 3/1/2013 10:06 AM, Adam Goryachev wrote: > ... >>>> Is it worth reducing the chunk size from 64k down to 16k or even smaller? >>> >>> 64KB chunks should be fine here. Any gains with a smaller chunk would >>> be small, and would pale in comparison to the amount of PITA required to >>> redo the array and everything currently sitting atop it. Remember you'd >>> have to destroy it and start over. You can't change chunk size of an >>> existing array. >> >> Actually, you can. For levels 0 and 4,5,6. > > First, I'll reiterate that a smaller chunk size likely is not going to > yield real workload gains for Adam. And it obviously would decrease his > FIO numbers, making him think performance decreased, even if it actually > increased slightly with his real workload. No contest. I have effectively no experience at these hardware performance levels. > Speaking strictly now from a knowledge transfer standpoint, does this > chunk size change feature go back a ways or does it require a fairly > recent kernel and/or mdadm? Are there any prerequisites or special > considerations different from any other reshape operation? >From the announcement for mdadm version 3.1, October 2009: > It contains significant feature enhancements over 3.0.x > > The brief change log is: > - Support --grow to change the layout of RAID4/5/6 > - Support --grow to change the chunksize of raid 4/5/6 > - Support --grow to change level from RAID1 -> RAID5 -> RAID6 and > back. > - Support --grow to reduce the number of devices in RAID4/5/6. > - Support restart of these grow options which assembling an array > which is partially grown. > - Assorted tests of this code, and of different RAID6 layouts. > > Note that a 2.6.31 or later is needed to have access to these. > Reducing devices in a RAID4/5/6 requires 2.6.32. > Changing RAID5 to RAID1 requires 2.6.33. So I'm sure there are plenty of older systems that can't do this, but current distros should all have it. Phil