Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Optimal RAID configurations for relational databases

Re: Optimal RAID configurations for relational databases

From: Malcolm Blackhall <m.blackhall_at_pbnec.com>
Date: 1998/12/09
Message-ID: <366F76CE.CC3CA1EF@pbnec.com>#1/1

Yes, you will take a performance hit using RAID 5 for two reasons. The first is the very nature of RAID 5. It maintains parity as part of each stripe to provide redundancy. Whenever you write to the device, the controller has to calculate parity and write it as well as the data. On the up side, you can get some read performance improvement under certain circumstances if the system is tuned properly.

The second reason is that with RAID 5, as far as the database server is concerned, you have a single drive. Both MS SQL Server and Oracle perform better if you can separate certain things on different drives and controller channels.

There is another issue with the RAID 5 configuration you describe. The Smart 2DH controller is a two channel device. For optimal performance, you will put two drives on one channel, and three on the other channel. RAID 5 will protect you in the event of the failure of any one of these drives, however, if you lose one of the controller channels, you are out of business because you have lost either two or three of your drives. My experience is that drive failures are much more common than controller channel failures, but controller channel failures do happen. If it is really important to stay up in the event of such a channel failure, you either need a real disk array where you can have each of the drives in the RAID 5 set on separate internal channels, or you need to use the hardware you have differently.

Personnally, I would try to get another drive and configure the six drives as three RAID 1 devices. You get a little less storage and pay a little more, but get some peace of mind for it. Received on Wed Dec 09 1998 - 00:00:00 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US