From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Eric Wood" Subject: Re: How safe is software RAID compared to how safe hardware RAID is!? Date: Fri, 21 Nov 2003 09:59:30 -0500 Sender: linux-raid-owner@vger.kernel.org Message-ID: <00a901c3b040$10e451e0$9100000a@intgrp.com> References: <16314.40792.616508.878124@notabene.cse.unsw.edu.au> <16314.44161.438158.27514@notabene.cse.unsw.edu.au> <012e01c3ae33$6cde29d0$cc07a8c0@neptune> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: To: linux-raid@vger.kernel.org List-Id: linux-raid.ids 3tcdgwg3 wrote: > Did I do anything wrong? > > I didn't see RAID 1 hits multiple drives concurrently, specially for > reading. > > I set up RAID 5, RAID 1 and single disk partitions on x86 machine, > kernel 2.4-18. > > For 800 MB breading, RAID 5 take 2/3 of time used by single disk, > but RAID 1 take 1.1 of time used by single disk. Doing anything (reading or writing) with RAID 1 set will always be slower than with a single drive. Because even with reading, you're still updating the atime on both drives. As a test, mount that partition with noatime and see if RAID-1 reading test nears to 1. Here's a nice message from while back: http://www.cs.helsinki.fi/linux/linux-kernel/2002-17/0313.html Jaime Medrano wrote: > I have taken a look at the read balancing code at raid1.c and I have found > that when a sequential read happens no balancing is done, and so all the > reading is done from only one of the mirrors while the others are iddle So RAID-1 sets don't stripe read. RAID 5 is forced to stripe seek because of its very nature of having XORed data across the drives. -eric wood