> -----Original Message-----
> From: Mark D. Stock [mailto:mdstock@mydas.freeserve.co.uk]
> Sent: Friday, September 15, 2000 4:32 PM
> To: kagel@bloomberg.net
> Cc: informix-list@iiug.org
> Subject: TBLSPACE_STATS to turn off/on?
>
>
> "Art S. Kagel" wrote:
> >
> > You will indeed lose the onstat -g ppf output but the
> overhead associated
> > with TBLSPACE_STATS is not insignificant. I have NEVER had
> this parameter
> > set through the four versions that have supported it and
> have never had any
> > memory corruption or other problems that were traced to
> this facility.
> >
> > My recommendation? Keep it off unless you suspect some
> bottleneck then
> > turn it on for a reasonable period, say a week to gather
> stats then off
> > again after analysis.
>
> Mmmm, I'll throw my 2c in here. I NEVER turn off
> TBLSPACE_STATS, because
> I use the stats a lot. Also I seem to work with a lot of 24 hour
> operations, where asking to quickly bounce the engine to change some
> parameters is met with looks of confusion and comments like,
> "didn't we
> already do that last year?" When it comes to diagnosing performance
> issues, I like to have those stats available, even at the cost of a
> little performance.
>
Same here. I also use those stats to take snapshots of the system for
various maintenance reports and such, like my 'hot tables' report and my
'reorg planner' report. Makes life much better in the long run. (Besides,
if performance really suffers that much, I would temporarily unset
TBLSPACE_STATS and gain a performance boost. Really helps the 'miracle
worker' persona . . . . 8-)
> I REALLY look forward to the day when ALL parameters can be changed on
> the fly. :-)
Hear, hear <clap, clap, clap, clap>