From owner-freebsd-questions@FreeBSD.ORG Thu Jun 30 21:51:20 2005 Return-Path: X-Original-To: questions@freebsd.org Delivered-To: freebsd-questions@FreeBSD.ORG Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 70D2F16A41C for ; Thu, 30 Jun 2005 21:51:20 +0000 (GMT) (envelope-from chad@shire.net) Received: from hobbiton.shire.net (hobbiton.shire.net [166.70.252.250]) by mx1.FreeBSD.org (Postfix) with ESMTP id 4462A43D1F for ; Thu, 30 Jun 2005 21:51:20 +0000 (GMT) (envelope-from chad@shire.net) Received: from [67.161.222.227] (helo=[192.168.99.68]) by hobbiton.shire.net with esmtpa (Exim 4.51) id 1Do6wI-0009bn-RF for questions@freebsd.org; Thu, 30 Jun 2005 15:51:19 -0600 Mime-Version: 1.0 (Apple Message framework v730) In-Reply-To: <20050630213413.GJ2392@dan.emsphone.com> References: <42C45161.1070402@toldme.com> <20050630204448.8E0F543D4C@mx1.FreeBSD.org> <20050630211028.GP1280@rabbit> <20050630213413.GJ2392@dan.emsphone.com> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <87C2B100-4BF9-4001-B9A0-FD2E8312F00A@shire.net> Content-Transfer-Encoding: 7bit From: "Chad Leigh -- Shire.Net LLC" Date: Thu, 30 Jun 2005 15:51:18 -0600 To: FreeBSD Questions X-Mailer: Apple Mail (2.730) X-SA-Exim-Connect-IP: 67.161.222.227 X-SA-Exim-Mail-From: chad@shire.net X-Spam-Checker-Version: SpamAssassin 3.0.3 (2005-04-27) on hobbiton.shire.net X-Spam-Level: X-Spam-Status: No, score=-0.2 required=5.0 tests=AWL,BAYES_50, GREYLIST_ISWHITE autolearn=disabled version=3.0.3 X-SA-Exim-Version: 4.2 (built Mon May 30 00:43:02 MDT 2005) X-SA-Exim-Scanned: Yes (on hobbiton.shire.net) Cc: Subject: Re: RAID Cards X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 30 Jun 2005 21:51:20 -0000 On Jun 30, 2005, at 3:34 PM, Dan Nelson wrote: > In the last episode (Jun 30), Mark Bucciarelli said: > >> On Thu, Jun 30, 2005 at 04:48:18PM -0400, Simon wrote: >> >>> Just because there is no monitoring tool available due to lack of >>> support, doesn't mean the card itself is bad. I much prefer >>> hardware implementation than software. True hardware RAID frees up >>> a lot of CPU time if you have heavy IO and software just can't keep >>> up if you utilize CPU intensive apps. >>> >> >> Why do you say hardware raid frees up "a lot of CPU time?" Have you >> measured this? >> >> Do you have any servers that are cpu-bound instead of io-bound? >> >> I am having this exact discussion with my business partner at the >> moment--he is also a proponent of hardware raid. I don't see the big >> win in hardware raid. >> > > The three big plusses for hardware raid are: if you get one with > battery-backed cache (strongly recommended), then the array can cache > raid-5 writes until it gets full stripes, and can hold off doing > mirror > writes if there are pending read requests. Also, if your power goes > out or the system spontaneously reboots, you won't have to rebuild > parity or resync the mirrors (assuming battery-backed cache). And > finally, hardware raid cards will automatically rebuild onto a hot > spare if available and you can swap out the dead drive and swap a new > spare in without having to run a single command. I am not an expert at all, but I believe the following to be true and advantages of true HW raid cards. To add to the above from Dan Nelson. -- HW raid cards reduce the traffic on your PCI bus. One read or write request is issued and one set of data goes over the PCI bus. The card itself worries about talking to the drives and reading or writing the data from the appropriate drives -- even if you are not CPU bound in terms of fully using the complete CPU, if you are busy, the CPU has a queue of things to do and I like to keep the CPU queue as small as possible... For example, busy PHP based sites can queue up lots of processes even if the load does not peg the CPU due to other considerations, we can avoid extraneous context switches and extra CPU stuff -- good HW raid cards will have monitoring SW -- Adaptec, 3ware, and others do. -- simpler interface for the OS. The OS treats it as just another disk and so bugs in the OS (in your disk driver and RAID sw) don't corrupt your data as easily and in fact make you less OS and HW versions dependent, not more. --- Chad Leigh -- Shire.Net LLC Your Web App and Email hosting provider chad@shire.net