Comments on: Analyze This – 2

By: I Wish « Oracle Scratchpad

I Wish « Oracle Scratchpad — Fri, 16 Dec 2011 18:32:42 +0000

[...] Another difference between analyze and dbms_stats Anomalies with counting continued row activity Share this:TwitterLike this:LikeBe the first to like this post. Leave a Comment [...]

By: Blogroll Report 26/03 /2010 – 02/04/2010 « Coskan’s Approach to Oracle

Sun, 23 May 2010 22:21:38 +0000

[...] 6-Why to use DBMS_STATS instead of analyze ? Jonathan Lewis-Analyze this Jonathan Lewis-Analyze this 2 [...]

By: coskan

coskan — Wed, 05 May 2010 08:40:22 +0000

I did another test

1- disable the _optimizer_compute_index_stats by setting false
2- create the index
3- gather stats with analyze index compute statistics
4- drop index
5- create index again
6- gather statistics with dbms_stats.gather_index_stats(username,indexname)

What I see on trace is dbms_stats using different sqls then compute stats but populate dba_indexes with same values at the end. Documentation of dbms_stats also says “estimate_percent : Percentage of rows to estimate (NULL means compute).”

I think dbms_stats might differ when it comes to partitioned indexes but it needs more testing which I am not able to do now.

By: Timur Akhmadeev

Timur Akhmadeev — Wed, 05 May 2010 08:33:04 +0000

I think it’s best to ask Oracle support for clarification of what parameters are used for gathering index statistics automatically on CREATE INDEX. I don’t know, but I’d guess it is probably the same as calling gather_index_stats without additional parameters specified. I prefer not to trust automatic statistics gathering and run dbms_stats.gather_table_stats with cascade=>true and parameters I want such as no_invalidate=>false, etc.

By: coskan

coskan — Wed, 05 May 2010 07:49:55 +0000

Timur,

Actually that change is my reason to ask. What I want to be sure is if compute stats any different than dbms_stats.gathar_index_stats and if it is which one is used default for index creation ?

By: Timur Akhmadeev

Timur Akhmadeev — Wed, 05 May 2010 06:46:51 +0000

Starting with 10g CREATE INDEX automatically collects optimizer statistics. The behavior is controlled with _optimizer_compute_index_stats parameter which is true by default in 10g.

By: coskan

coskan — Tue, 04 May 2010 23:08:47 +0000

Jonathan

Is there a difference between (or is there any possibility second one can behave different) ;

create index ix_t1_n1 on t1(n1)

and

create index ix_t1_n1 on t1(n1) compute statistics

I could not find additional/different work in 10046 trace for both operation and also data in user_indexes are same.

By: Jonathan Lewis

Jonathan Lewis — Sat, 03 Apr 2010 10:25:53 +0000

Martin, Glad to hear the results of your poll. (And sorry to read about the impact of your presentation on stats)

By: mwidlake

mwidlake — Fri, 02 Apr 2010 12:55:20 +0000

Nice post about that difference.

Just some feedback for you Jonathan. It just so happens I was presenting on “gathering stats” at a UKOUG SIG earlier this week and I asked the audience some questions, including who still use ANALYZE.

The audience was small, only 25 or so people, but of them 2 still used Oracle 9 as their main DB, about 4 or 5 used V10.1 and everyone else used 10.2, so about 70%. No one used 11 yet as their main platform but about 50% had some 11g systems.

As for gathering stats, no one was using ANALYZE anymore, which is the first time I’ve found no one still using ANALYZE as a matter of course. Even the V9 people had swapped to DBMS_STATS I am glad to say.

Several of us did admit to using ANALYZE for “quick tests” though :-)

Martin