Re: dbms_utility.analyze_schema('SYS', 'COMPUTE')

In article <7324b7$ie3$1_at_juliana.sprynet.com>, vic04_at_sprynet.com says...
> Interesting. So, to paraphrase, this means that as long as ANALYZE is
> not run against fet$ or uet$, then no deadlock or corruption should occur?
> In practice, I am actually only interested in updating the stats for the
> following tables and their indexes:
> CCOL$
> CDEF$
> COL$
> CON$
> OBJ$
> USER$
>
> ----
> Vic Thomas
> Database Administrator
> Associated Software Consultants
> vic04_at_sprynet.com; vic_at_asconline.com
>
>
>
> Jonathan Lewis wrote in message <01be13e6$73dfdf90$0300a8c0_at_WORKSTATION>...
> >
> >
> >Rich Woods <rawoods_at_concentric.net> wrote in article
> ><36536E21.454967D6_at_concentric.net>...
> >> Vic,
> >>
> >> As the Analyze command updates the data dictionary with information
> >extracted
> >> from the specified table/index/schema, a deadlock will most likely
> >occur when
> >> the Analyze command attempts to gather this information from the
> >data dictionary
> >> table where statistics are stored.
> >>
> >
> >Specifically - when you analyze a table you may need to perform some
> >space
> >management operations for the sorts etc; i.e. inserts/deletes on
> >fet$ and uet$,
> >so if analyze these two tables and they are large, you could hit a
> >deadlock.

I seem to remember from somewhere Oracle saying that the tables in the data dictionary were built to provide the best performance using the rule based optimizer and that generating statistics for the tables could cause a switch to cost based optimization and a performance bottleneck. Perhaps this isn't true any longer. Does anyone know for sure?

Bob

--

Bob Withers                        Do or do not, there is no try
bwit_at_pobox.com                                           Yoda
-------------------------------------------------------------------------