Re: How to measure sort time

"Gabriel Gonzalez" <no-spam_at_no-spam.com> wrote in message news:<wKKdnanID6-_L-qjXTWchQ_at_giganews.com>...
> > steps in terms of memory usage, whereas I'm interested in sheer time, with
> > the maximum precision.
>
> Hmm... Whatever tool you use you might get high precision, but not high
> accuracy. There are many things that can make a database take longer or
> shorter to run the same exact query. If your benchmark thingie tells you a
> query runs in 11.546 seconds, that might be true, but the next time you run
> the query it will probably run in a somewhat different time.
>
> In other words, if you need a ball-park time for your query, using your
> watch's seonds hand will be enough...

I'd examine v$sesstat and examine the entries for that session - join to v$statname and v$session - look at entries that have v$statname.name like '%sort%.
Go by how many sort operations and how many bytes.

if you're interested in more detail, then look into setting event 10046 and generate some trace files with waits=>true. there exists much literature out there for howtos - http://www.hotsos.com has some excellent papers on the subject.

Paul Received on Tue Mar 18 2003 - 22:55:19 CST