Re: Database Performance Metrics - What to use?

buckeye234 <member13594_at_dbforums.com> wrote in message news:<2641160.1047585596_at_dbforums.com>...
>
> I was asked to provide generic KPI's, but I responded that they would
> vary by database type. I just wanted to validate my thoughts
> (sometimes those thoughts stray from the realm of normal ;). It looks
> like it is reasonable to present different KPI's for databases that
> serve different purposes.

In a sense, I'd say the comment about 'Dilbert world' was spot-on... IF management don't know what they want to measure but they are asking for metrics, it sounds like time to check who is the opposition vendor doing the rounds! :(

In the interest of fairness:
I'd go along with committs/number of users or unit of time. It's served me well in OLTP over the years.

Response time makes heaps sense but unless you have specialized software or the "hooks" built-in, you cannot accurately measure it. So, leave it out for the time being.

Do not underestimate uptime. It can be VERY impressive!

As for the imaging stuff: it's hard to get something for this. Given that it's mostly one-offs, I'd go with number of reads/user. Or better yet: number of Gbs read per user or unit of time. Can be quite amazing!

For the reports, number of reports/hour or /user, completed. Or if they are in jobs, number of jobs.

For things like Oracle Applications, the number of concurrent jobs finished per unit of time.

Do some graphics of load across a day and across a week. They look good and can be very impressive.

HTH
Cheers
Nuno Souto
wizofoz2k_at_yahoo.com.au.nospam Received on Thu Mar 13 2003 - 21:09:29 CST