From mboxrd@z Thu Jan 1 00:00:00 1970 From: "=?ISO-8859-15?Q?Stefan_/*St0fF*/_H=FCbner?=" Subject: Re: MD RAID Bug 7/15/12 Date: Tue, 02 Oct 2012 00:27:56 +0200 Message-ID: <506A18EC.1040602@gmail.com> References: <022FFCA9-5D0E-41DB-9617-202C6FAADD06@rightthisminute.com> <5068B4BF.6000807@gmail.com> Reply-To: st0ff@npl.de Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: Sender: linux-raid-owner@vger.kernel.org To: Chris Murphy Cc: Linux RAID List-Id: linux-raid.ids Am 01.10.2012 00:16, schrieb Chris Murphy: >=20 > On Sep 30, 2012, at 3:08 PM, Stefan /*St0fF*/ H=FCbner wrote: >=20 >=20 >> Also off topic: 12 drives would be as "nearly unalignable" as 19 are= =2E >=20 > I'm not sure what you mean by unalignable. A separate question is if = a 4K chunks size is a good idea, even with 24 disks, but I'm unsure of = the usage and workload. >=20 >=20 [...] By that I mean that optimal alignment has to do with a reasonable chunk size and an amount of data disks that is/should be a power of two. If you made a databases commit chunk size 64k, you'd create maybe a raid6 of 6 drives with a 16k chunk size. That way you'd spare many many read-modify-write operations, as on every commit it'd write all 6 disks once and no rmw would happen. You'd also have to take care of proper partition placement. That probably is what you know as alignment. I just want to make it more concious that not only the placement of a data partition goes into alignment, also the size of a "normal" data packet relative to chunk size and amount of disks used. cheers, stefan -- To unsubscribe from this list: send the line "unsubscribe linux-raid" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html