This is what it buys you:

1. I/O is better spread across multiple drive heads.  Not so
important for writes, if you have a good-sized write-cache, but
especially helpful for reads.  TSM does not do round-robin allocation
of DB pages across it's volumes.  It fills up one, then works on the
next.  Even if you have gotten to the point where you have multiple
dbvols, you'll tend to allocate pages for a node together on the same
dbvol at first.  Then when you go to reference them, you'll be
hitting one drive more than the others.  RAID5 will do a reasonable
job of spreading things across multiple heads.

2. FAStTs have a nice feature where they will automatically replace a
failed drive with a hot spare.  Then you can replace the hot-spare
drive at your leisure.  Without this, if you were using single LUNs
and one failed, you would have to replace the failed drive, then
re-mirror within TSM.  Not nearly as automatic and your window of
exposure will be greater.  In our shop, we have separated management
of the FAStT devices from management of the TSM environment (i.e.,
different staff), so it is particularly attractive to the folks who
manage the FAStT boxes to just have hot-spare coverage for
everything.  Then they don't have to worry about the impact on TSM
when a drive fails.

At 03:40 PM 1/20/2006, Jack Coats wrote:

Just curious, and playing a little devils advocate, but does it buy
anything to mirror over raid5?  Wouldn't it be more efficient (processor
on the san, and storage) just to put volumes out there and let TSM
manage them?  You would want to have the mirrored volumes on separate
physical drives (preferably in different trays, etc) also.

--
Paul Zarnowski                            Ph: 607-255-4757
Manager, Storage Systems                  Fx: 607-255-8521
719 Rhodes Hall, Ithaca, NY 14853-3801    Em: psz1 AT cornell DOT edu