I think, as usual, IT DEPENDS ON YOUR ENVIRONMENT.

The RAID5 prevents outage due to a disk failure.
But there is some exposure to other failures.

Your RAID5 controller can fail.  And there is always that traditional
bugaboo of HUMAN failure - I always worry about some enthusiastic trainee or
accident-prone vendor tech accidentally reconfiguring the RAID set while I'm
not arround.

There is also some, very LOW likelihood of a LOGICAL (software) error
writing to the DB or log.  Tivoli has stated in the past, that there are
cases where they can detect a bad write to the log and STOP before
replicating the bad write to the mirror copy (I've never seen this occur
myself, so I don't know how realistic this is), which is why they recommend
using TSM mirroring instead of OS mirroring.

So, I think you have to evaluate your situation; what is your exposure; what
is your tolerance to an outage, and what is your tolerance for data loss?
Do you want to have to explain ANY downtime  / data loss to your management?

In systems where the transaction load will allow it, I put one copy of the
DB on RAID5, put the log in ROLLFORWARD mode, then mirror just the LOG to an
internal disk (to avoid the WRITE penalty), or to a RAID5 device on a
different adapter.  That way, EVEN IF you lose the RAID5 set, you still have
the log, which means you can always restore the DB from a DB backup tape,
and roll forward to current; the only thing you lose is some time.  So you
achieve almost bullet-proof coverage with not much cost in space.

If your transaction load is too high to stand the overhead of ROLLFORWARD
mode, and you use NORMAL mode, then if you have a failure you lose your
transactions back to the last DB backup.  That's OK for a lot of sites.  If
all you are backing up is some servers under your control, and all you lose
is your previous night's backups, then fine.

But,  if you have a lot of far-flung clients not under your control, or if
you do a lot of archiving it gets trickier, 'cause you have to track down
what archives were lost and do them again.

And if you are in an HSM environment BOW HOWDY, then you can't afford ANY
outage, even TIME.

So, you have to think about

1) What is your exposure - do YOU understand exactly what you would lose if
you had an outage?
3) What are you willing to pay to cover your exposure?
4) Does your management understand your exposure?

and set yourself up accordingly.

Just my 2 cents.

(and by the way, does our 2 cents worth stay constant, or do we have to
pro-rate it for inflation from year to year?)

Wanda Prather