Post reply

2 mirrored drives or 6 smaller RAID 10 drives...????

  • February 24, 2005 at 11:32 am

    #164610

     What would be better performance wise for a 4GB SQL 2000 database....

    2x146GB mirrored that would contain the DB datafiles

    or

    6x36GB RAID 10 that would contain the DB datafiles

    Yes, I realize the size of the drives is a bit of overkill but it happens to be what we have lying around.

  • February 24, 2005 at 12:21 pm

    #543988

    as a general rule of thumb, the more drives you can stripe your files accross the better your performance will be. Since in the first scenario your database files will be mirrored you are actually only using one drive for performance. The second scenario will give better performance.

  • February 24, 2005 at 12:23 pm

    #543989

    Great thanks....can i get one more opinion?

    the database is quite small so i'm thinking that the disk size really is overkill....if i used 4x36Gb RAID 10 would that be pointless?

  • February 25, 2005 at 3:49 am

    #544108

    If you already have the disks, why not use them? Yes, it is overkill, but the DB will probably grow, and also you may need to add other databases later. We started with 20 GB and today (after 3 years) the size of our main database is over 150 GB. Anyway, use RAID 10 and disks with high speed - that affects performance a lot. Mirroring is not a good option for the disk with DB files in most cases.

  • February 25, 2005 at 5:15 am

    #544119

    Here you can find some links, that might help (don't worry, they are all in english )

    http://www.insidesql.de/component/option,com_bookmarks/Itemid,42/catid,26/

    --
    Frank Kalis
    Microsoft SQL Server MVP
    Webmaster: http://www.insidesql.org/blogs
    My blog: http://www.insidesql.org/blogs/frankkalis/[/url]

  • February 27, 2005 at 11:00 am

    #544328

    I have not had experience with Raid 10. In general, more spindles are better, but more logical drives are also good. I would suggest arranging your 6 36GB drives in three Raid 1 pairs. This would allow data files on one pair, log files on another pair, and tempdb on the third pair. Place data and log backups on different drives from their respective online files, and place indexes on the tempdb pair, unless you can afford a fourth pair. This assumes a fair proportion of write activity - if your db is mostly read-only, separating the log files is less of a priority.

    I would be interested in hearing from other readers about the advantages of striping vs. the disadvantages of a limited number of logical drives - to me striping makes more sense for a database server when you have LOTS of spindles and can enjoy the best of both types of IO distribution. Any comments?

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply