Re: Please recommend a disk configuration for the server

Chris Pitzel wrote in message <361A8727.3D086DC4_at_nospam.usask.ca>...
>satar_at_my-dejanews.com wrote:
>>
>> What's wrong with 9GB drives? we don't know his situation, he might need
 9GB
>> drives because his APPLICATION_DATA Tablespace might/will be 6GB. I am
 not a
>> NT system Admin, so I cannot comment on NT's mirroring.
>
>A single 9 gig drive cannot support nearly the same I/O load as a pair
>of 3 4.5gig drives in RAID 5. Depending on the specific application (in
>this case, the Oracle database), and load on the database, a single 9
>gig drive may have trouble providing a satisfactory number of I/O
>operations per second versus a RAID 5 array of 3 4.5gig drives which
>should fare better.

True, but I'd say *only on reads*. An uncached RAID5 array would generally be much slower on writes. And I would want a cache that was battery-backed and bullet-proof before using it for a db application - and most aren't. The only way you'd ever really know is to do some proper performance testing. Still, always worth remembering the write penalty of RAID5, especially when you think about the placement of redo, rollback and temp (and data and index, depending on your application).

>
>It's difficult (in other words, impossible) to determine whether or not
>this is the case from the original post.

True.

> However, as drives are
>continually increasing in size (you can't even buy 2 gig drives
>anymore), the I/O rates of drives are only increasing slightly.

Yep, it's a bit of a downer when you can't buy some smallish drives to dedicate to your control files and redo logs (and a few heavily used, yet smallish tablespaces you'd like to pull away on their own). Same thing when setting up some stripe sets for max I/O - the disks are so big theses days you end up with massive stripe sets even with just 2 or 3 drives. Damn progress! :-)

MotoX. Received on Wed Oct 07 1998 - 00:00:00 CDT