From mboxrd@z Thu Jan 1 00:00:00 1970 From: Roger Heflin Subject: Re: "raid" versions of hard drives for software raid? Date: Thu, 03 Jul 2008 11:06:54 -0500 Message-ID: <486CF91E.4090705@gmail.com> References: <873amsqbag.fsf@uwo.ca> <486BE720.3090507@redhat.com> <87tzf8ovll.fsf@uwo.ca> <87abgzvxxn.fsf@uwo.ca> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <87abgzvxxn.fsf@uwo.ca> Sender: linux-raid-owner@vger.kernel.org To: Dan Christensen Cc: linux-raid@vger.kernel.org List-Id: linux-raid.ids Dan Christensen wrote: > Dan Christensen writes: > >>> Dan Christensen wrote: >>>> I'm looking at either the Western Digital 500G RE2 drives or the cheaper >>>> 500G SE16 drives. I have read that for use with a hardware raid card, >>>> the RE2 drives are more appropriate, and I'm wondering if the same is >>>> true for software raid. >> The difference that is most commonly described, and that I should have >> highlighted, is TLER: Time-Limited Error Recovery. Apparently, the SE16 >> drives can take a long time to recover from an error (up to two minutes, >> I believe), and hardware raid controllers can kick the drives out of the >> array when it would instead be better for the drive to return a >> read/write error and let the raid controller deal with it. >> >> My question is really whether this logic applies to linux software raid. > > One more reference: > > http://en.wikipedia.org/wiki/Time-Limited_Error_Recovery > > It certainly sounds to me like TLER is also appropriate for software > raid, so I'm going to go ahead and get the RE2 drives. > > Dan If you were trying to salvage as much data as possible off of a drive TLER would also be nice, since it would lower the time to get a bad sector error. I don't really understand how much use it is to try that many times, the drive should be able to try 1x per rev, so at 7200, 120 times per second, I would wonder how many times that they get a good read after having failed the first 120 times, much less after the first 7 seconds (960 failures) of trying, or even longer periods without TLER. In fact I would think for a RAID drive one would want even lower than 7 seconds. Roger