Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Global/partition level statistics

Global/partition level statistics

From: Parveen Ahuja <pahuja_at_mitre.org>
Date: Thu, 09 May 2002 14:16:06 -0400
Message-ID: <3CDABCE6.69699661@mitre.org>


Hi,

I need some help on how to gather the database statistcs ?

I have close to 250 GB of database spread into 150 partitions, each partitions is close to 1.5 to 2 GB. And each partition has 16 tables. These partitions are locally managed partitions and partitions are based on column ID and it is RANGE partitioned.

Example:
Column ID Range

101 - 200
201 - 300
301 - 400

In case of only PARTITION LEVEL STATISTICS queries like below had a very good performance

/* Hitting only 1 partition */
SELECT col1, col2
FROM tableA
WHERE ID > 100 and ID < 200

while the following query had a bad performace.

/* Hitting more than 1 partition */
SELECT col1, col2
FROM tableA
WHERE ID > 100 and ID < 300

In case of GLOBAL STATISTICS + PARTITION LEVEL STATISTICS, the above performance almost reversed. I get good performace with queries hitting more than 1 partition, while poor performace when queries are hitting only one partition.

What strategy to used to get the best performace in both the scenerios ?

Other facts:
Statistics is gathered using the DBMS_STATS.GATHER_TABLE_STATS... with CASCASE, AUTO SIZE etc.

Any Help...
Thanks in Advance.


Received on Thu May 09 2002 - 13:16:06 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US