Re: Benchmarks

mbrugo_at_yahoo.it wrote:

> On Mon, 13 Jan 2003 14:47:21 -0000, Norman Dunbar
> <Norman.Dunbar_at_lfs.co.uk> wrote:
>
> >I don't know about benchmarking these bits of kit, but I'd be asking
> >myself 'what experience do I have in house for administering hardware'
> >and working up from there. For example, if I had a full compliment of
> >Microsoft experience, I'd probably be loath to move over to a Unix
> >installation - regardless of how much better is is/would be because I'd
> >need to shell out vast sums of money to retrain staff or employ staff
> >who know Unix.
> >
> >Just a thought.
> >
> >Cheers,
> >Norman.
> >
> >
>
> Dear Norman
> our problem is not "The experience"
> We have skill to administer all this Hw.
> We want to buy the best Hw for Oracle.
> we have find some result on tpc.org but we need more benchmarks.
>
> Mike

Seems inconsistent that you have the hardware expertise and are asking the question but here's my go at it.

1. Oracle releases bug fixes for Solaris first, HP second, others later.
2. During hard crashes HP tends to be more likely to corrupt the kernel than Solaris.
3. It doesn't much matter. The differences are not significant enough to care about. TCP benchmarks aren't worth the paper they are scribbled on. They are completely meaningless except to the PR marketing types that pay for them. What you are doing will never be exactly what anyone else is doing. Your tuning and code will never be the same as anyone else's. Go for a contract based on your ability to maintain the systems and keep them running 7x24.

And, just as an aside, I find it nearly impossible to believe that you folks have identical expertise with Solaris with HP/UX and with AIX. I'd seriously consider the cost of having an O/S you don't know as well vs. one you do know well.

Daniel Morgan Received on Mon Jan 13 2003 - 11:54:16 CST