Re: raid level

Mladen Gogala wrote:
> On Fri, 16 Dec 2005 10:44:11 -0800, DA Morgan wrote:
>
> > Throwing hardware at a problem, at least to me, is the sign of an
> > unskilled technician.
> aka "manager". They're the ones with the check signing power, in case
> you've forgotten.
>

Not just that. "Refactoring" (newspeak for "redeveloping" or "redesigning")
is often much more expensive than adding more hardware. A bad app, like Sybrand said, will always be a bad app. The cost of re-doing
it properly will most likely involve a lot more than just a dba. dba's can't
fix a bad design without some major code surgery. That opens up a BIG can of worms of hidden costs!

Like for example: who's gonna test the new stuff and for how much? Who's gonna pay for the downtime while the new one is deployed? How much will it cost to move data to the new one? What about training users? Documentation? Help files? Processes? etcetc...

What about re-doing all the file systems for moving to a raid10? Optimizing
the OS to make best use of it? Patch up Oracle to use all the trimmings
of the given solution (typically aio and/or dio) *and* TEST those patches?
Ah yes: Oracle is now bugless in 10gr2? Yeah! right...

More. The darn raid10 things are not as flexible configuration-wise as a
SAN: who pays for the design/test time needed to establish the best topology/features/options? Who pays for the systems and resources tied

up in testing everything BEFORE going into production?

Add the costs all up and the ubiquitous "slap on a SAN" might well be the most cost-effective solution! Sure: raid10 will always provide faster
throughput. But if all one wants is to reduce overnight batch processing by
25% while allowing for 50% size growth, then it is quite possible, feasible
and cost-effective to do the hardware shuffle with a SAN.

These things are not black on white: multiple shades of grey are the order of the day... Received on Mon Dec 19 2005 - 23:18:19 CST