Re: To RAID or not to RAID (...or how to RAID)

"Anton Buijs" <aammbuijs_at_xs4all.nl> wrote in message news:<alqlmp$40t$1_at_news1.xs4all.nl>...

> I've read those 3 documents a few times now. I still don't understand the
> objections of ixora and quest.
> Why spent a lot of time (and downtime) in moving datafiles and tablespaces
> to tune disk I/O? The heaviest loaded disks could be a bottleneck and I can
> improve that by making more disk spindles available - at the cost of more
> disks when the space is not really needed.

It's all relative, I suppose. Well over 90% of sites out there with a performance problem start with an I/O tuning and balancing exercise. Without even checking if they have an I/O problem to begin with. Result: if they use SAME or not is immaterial and makes no diff whatsoever. Too much other overhead for SAME to be relevant.

Most of the problems have to do either with horrible SQL, horrible index design and/or horrible optimizer bugs creeping in. Solving these causes is where most consultants spend their time. I/O is nowhere to be seen in most cases.

Now, both ixora and quest usually deal with databases after these initial tuning steps have been taken. As a result, in their case I/O may well be a relevant factor. In which case indeed the comments against SAME might well be warranted. If you are after the last 10% of performance out of a system.

If not, then the ease of config and use of SAME is perfectly OK and should not result in any major problems. If I/O is not the najor problem to start with, then SAME is valid. It is a disk space management technique more than it is a performance technique.

Cheers
Nuno Souto
nsouto_at_optushome.com.au Received on Thu Sep 12 2002 - 22:17:59 CDT