From mboxrd@z Thu Jan 1 00:00:00 1970 From: Neil Brown Subject: Re: Chunk size for SSD Date: Wed, 3 Mar 2010 07:22:07 +1100 Message-ID: <20100303072207.5d698723@notabene.brown> References: <20100302190843.GA2446@lazy.lzy> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20100302190843.GA2446@lazy.lzy> Sender: linux-raid-owner@vger.kernel.org To: Piergiorgio Sartor Cc: linux-raid@vger.kernel.org List-Id: linux-raid.ids On Tue, 2 Mar 2010 20:08:43 +0100 Piergiorgio Sartor wrote: > Hi all, > > I know the topic might have already been discussed, > but some points are still unclear to me. > > Assuming to build a RAID-10 with two SSD, someone > mentioned that the proper chunk size should be > the erase block size, usually 128K. > > Are these 128K a guessed value or there is a way > to know the erase block size from the device itself? > > Second question, assuming metadata 1.1 is used, do > the chunks start with a chunk-size aligment? > I mean, in a RAID-10, with 1.1 metadata, does the > effective storage area start at a multiple of 128K, > if a chunk of 128K is used? No - with 1.1 and 1.2 metadata there are no guarantees about alignment of the start of the data section. That is something that I plan to fix in 3.1.2. > > BTW, what if there is a partition under the RAID, > this will offset (how much?) the start of the RAID. It depends on the tool you use the create the partition. Some enforce an alignment on cylinders, others give you complete control. I think 'sfdisk' is one of the most flexible. NeilBrown > > In general, would it be possible to specify an > alligment for the effective storage area of the RAID? > > Thanks a lot in advance, > > bye, >