Re: Oracle & RAID strategy question

"Bonjour" from Paris,

From the "Veritas Database Edition for Oracle" documentation (p-26) relative to the volume layout :

Create stripped volumes across about 4 to 10 disks only. Do not stripe across less than 4 disks.Try to stripe across disk controller. For sequential scans, do not stripe across too many disks or controllers. The single thread that processes sequential scan may not be able to keep up with the disk speed.

For moast workloads, use the default 64K stripe unit size for stripped volumes and 16K for Raid-5 volumes.

"Au revoir" from Paris
Hieraklion

Kate Kwiatek a écrit :

> Greetings....
>
> I have a Sun E450 to configure for some small Oracle databases that will
> reside on the system. The system has 8 36-Gig hard drives and 2
> controllers. For now, the Oracle databases are fairly small and not real
> write-intensive at all. One of the databases is the database for our
> Oracle Portal application and the other database will hold financial
> information for a third-party forecasting and reporting tool called
> Control8 (KCI Computing). I'm not the actual unix sys admin who will be
> configuring the box, but the Oracle DBA driving the general
> configuration request.
>
> My dilemma is that I'm torn between which implementation of software
> RAID i should have the sys admins build. We must have complete
> redundancy so 4 disks will definitely be mirrored -- an unfortunate
> overhead, but my company can afford it. The two choices I have are the
> following:
>
> 4 disks each individually mirrored and manually laying out my Oracle
> directories to separate data, indexes and database software...
>
> or
>
> 2 logical volumes(4 disks-1controller each) striped and mirrored with an
> LVM like Veritas Volume Manager
>
> The problem I have with the second option is that I cannot directly
> control the separation of data and indexes in the Oracle environment,
> but since the databases are small and not highly-transactional, I am
> thinking the disk i/o benefit from the striping alone will negate this
> issue.
>
> Does anyone have any advice on which setup I should go with?? I
> understand the implications of failures for each, so I'm really just
> wondering which would offer better read-write performance for my
> particular database environment.
>
> I appreciate all advice....much thanks to all who reply:)
>
> Sincerely,
>
> Kate Kwiatek
> Manager - Technical Control
> Raytheon Company
> kwiatek_at_raytheon.com
Received on Tue Apr 17 2001 - 07:30:30 CDT