Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.server -> Re: Table access slow
Felipe Mart?nez wrote:
>>Reading it all, some questions: >>- are you sure you did what you described, and just truncated >> the table (and not drop it - in which case you could miss indexes)?
>>- post explain plans - even if you cannot change the statements, it can >> point in a direction with a bit more accuracy than this stabbing in >> the dark. >>- Did you analyze the indexes as well? Correct me if I'm wrong, but a >> ROT was to analyze the index, and estimate the table. >>- Did you have statistics in the first place? 7.3 CBO was not up to >> the standard of 9i/10i (which I'm currently downloading...). >> In your test/acceptance environment, drop all stats and see if >> performance is back to 'normal'. Only then do so in production.
I
> recolected statistics before doing anything and I was confident it
> would work like many other times, so I lost those statistics. But I
> remember cache hitratio over 93%. Statistics I am collecting are from
> v$ tables which reset everytime the database starts up (don't they?)
>
>
> Thanks for your advise. I will post later with explain plan and
> indexes analyzed.
The question was: did you ever try without _any_ statistics? I've seen collecting statistics (and using the Cost Based Optimizer, which *will* kick in if optimizer_mode=choose, sometimes even with optimizer_mode=rule) with these early versions often do more bad than good.
And statistics meaning:
analyze table xxx estimate statistics;
analyze index xxxxy compute statistics;
See Oracle 7, release 7.3 Tuning mnl, part# A32537-1
-- Regards, Frank van BortelReceived on Tue Feb 10 2004 - 14:18:20 CST
![]() |
![]() |