| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> Re: Raid5 Vs Raid0+1 -- Raw Vs Solaris 9 Concurrent Direct IO UFS
Deen,
How can RAID-5 perform better than RAID-1 (especially 1+0) on reads? That's a new one I've not heard before. Oh, and writes? Don't worry about them -- we've got caching so it doesn't matter!
<rant>
Oh yes, the old "we've got so much cache it doesn't matter". Don't believe it: cache has to be managed (possible source of contention) and cache can be overwhelmed. At best, cache smoothes out the spikes, but it cannot help during a prolonged period of high activity.
I don't know why Sun thinks that there is a market for things like file-systems, but I've heard them waste hours trying to convince folks to use their wonderful new UFS (yawn!). At a large financial institution last May, I participated in an Oracle-based benchmark with four scenarios: "regular" UFS (i.e. no direct I/O), UFS with direct I/O, VxFS, and VxFS with "quick I/O" (now called ODM). Then, for kicks, the sys-admins and I added an "illegal" scenario -- "raw" logical volumes -- to the mix. The turd in the punch-bowl, the black sheep of the storage family, the fifth option that none of the vendors wanted to talk about. We boiled each of the benchmark results down to a single metric for relative comparison, with the first option (plain UFS) at a score of 22, VxFS with 18, UFS with direct I/O also at 17-18, VxFS with QIO/ODM at 12-13, and "raw" devices at 9. Of course, because the use of "raw" devices were "banned" from the competition (?), Veritas got their sale. Nev ermind we were using VxVM for the logical volumes... :-)
I count this as the best feat of marketing since Coke and Pepsi convinced everyone that "soft drinks" are a part of daily life, adding ump-teen layers of additional software to arrive a result that is measurably less than simply removing the whole kit-and-kaboodle. Paraphrasing the movie "Blazing Saddles": "File-systems? We don't need no stinking file-systems...", but often that attitude is rather like farting at the reception.
</rant mode="scuttles back into cave, pulls stone shut">
Hope this helps...
-Tim
Return-Path: <oracle-l-bounce_at_freelists.org> Received: from mail.sagelogix.com by ocs.sagelogix.com
with ESMTP id 32465831093458022; Wed, 25 Aug 2004 12:20:22 -0600 Received: by mail.sagelogix.com (Postfix, from userid 16)
id 2F445A85F9; Wed, 25 Aug 2004 12:12:44 -0600 (MDT) Received: from turing.freelists.org (freelists-180.iquest.net [206.53.239.180])
by mail.sagelogix.com (Postfix) with ESMTP id 4391BA85EA for <tim_at_sagelogix.com>; Wed, 25 Aug 2004 12:12:42 -0600 (MDT) Received: from localhost (localhost [127.0.0.1]) by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTP id 8405472EF75; Wed, 25 Aug 2004 13:17:47 -0500 (EST)Received: from turing.freelists.org ([127.0.0.1]) by localhost (turing [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 32243-30; Wed, 25 Aug 2004 13:17:47 -0500 (EST) Received: from turing (localhost [127.0.0.1])
by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTP id 1B21072F346; Wed, 25 Aug 2004 13:17:08 -0500 (EST)Received: with ECARTIS (v1.0.0; list oracle-l); Wed, 25 Aug 2004 13:15:38 -0500 (EST) X-Original-To: oracle-l_at_freelists.org
by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTP id 2B6BC72C66A for <oracle-l_at_freelists.org>; Wed, 25 Aug 2004 13:15:36 -0500 (EST)Received: from turing.freelists.org ([127.0.0.1]) by localhost (turing [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 30855-90 for <oracle-l_at_freelists.org>; Wed, 25 Aug 2004 13:15:36 -0500 (EST)
by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTP id 0473872DC48 for <oracle-l_at_freelists.org>; Wed, 25 Aug 2004 13:15:35 -0500 (EST) Received: from mail2av.state.nj.us (mail2av.state.nj.us [10.34.20.40]) by xmail2.state.nj.us (8.12.9/8.12.11) with SMTP id i7PIHVZs016591 for <oracle-l_at_freelists.org>; Wed, 25 Aug 2004 14:17:31 -0400 (EDT) Received: from revere.dol.state.nj.us(199.20.109.30) by mail2av.state.nj.us via csmap id 07825dac_f6c3_11d8_8c24_00304811e4b6_16026; Wed, 25 Aug 2004 14:17:31 -0400 (EDT)
X-Virus-Scanned: by amavisd-new at freelists.org X-archive-position: 8336 X-ecartis-version: Ecartis v1.0.0
X-list: oracle-l X-Virus-Scanned: by amavisd-new at freelists.org X-Spam-Checker-Version: SpamAssassin 2.63 (2004-01-11) on mail.sagelogix.com X-Spam-Status: No, hits=0.0 required=3.0 tests=none autolearn=no version=2.63 X-Spam-Level:
We are planning to move our current database to new Hardware Sun 12k + Hitachi SE 9970.
Our current database runs on Sun A5000 array with Raid0+1 on raw devices. Sun engineer who is here at our site for implementation of Hitachi SE9970 suggests that we should go Raid5 ( 3 + 1P ) in the parity group as RAID5 gives better read performance than raid0+1 and any writes are going to the cache any way, so we should not be worried about write performance. There is 16GB cache on the Hitachi. Database is going to be striped across 10 parity groups and each parity group consists of 4 physical disks.
Our application is about 60 to 70% reads most of the time, kind of a DSS and for few batch jobs it is 90% write and 10% read. Database is not that big just 200GB.
He also suggested us to use UFS instead of RAW as Solaris 9 has lot improvements in UFS especially with Concurrent Direct I/O and can perform better than RAW.
I am wondering anybody out there with real world experience with similar Hardware can throw some light on these 2 topics. If anybody can point any links or documents discussing these topics, highly appreciated.
Thanks
Deen
 |
 |