Re: Oracle and RAID

Ron Strouss wrote:
>
> "Glenn Stauffer" <stauffer_at_voicenet.com> wrote:
>
> >Dean Cunningham <deanc_at_wairc.govt.nz> wrote in article
> ><31E23307.6BFE_at_wairc.govt.nz>...
> >> kavis_at_usa.pipeline.com wrote:
> >> [snip]
> >> >
> >> > 3 and 5 negative on update/write operation, reliability is gained at
 the
> >> > expense of performance or need large cache,
> >> >
> >> > 0+1 best performance of all the RAID options, but doubles the # of
> >> > disks...so more $
> >>
> >> Fully agree. FYI Compaq were "quoting" a 15% drop in performance on
> >> their Array controllers with RAID 5 vs Raid 0/1.
> >> With their new controller they are "quoting" 5% drop.
> >>
> >> So perhaps the type of RAID controller can affect performance as
> >> well.....
> >>
 

> >Caching can make a big difference - almost enough to remove most of the
> >write penalty. My guess is that Compaq has either just added a much larger
> >cache to the controller or that and some other changes.
 

> >Glenn Stauffer
> >DBA
> >Swarthmore College
> >
>
> We're using RAID 5 with an HP 9000 ... I don't know who provided the
> RAID device, but it wasn't HP. I was thinking of moving the
> tablespace used for temporary tables to a non-RAID device as this
> tablespace has alot of write activity. Anyone try that and see if it
> makes a noticeable difference?
>
> Ron Strouss
> DBA
> Darigold, Inc.
> Seattle, WA

We had both RAID 0/1 and RAID 5. You get better performance out of RAID 0/1, because of the bottlenecking you get with RAID 5 setup.

JAY Received on Tue Jul 30 1996 - 00:00:00 CEST