Audit log select which can be a performance problem (IDM 6.0sp3)

Question

HI All,
we have in our production environment IDM 6.0sp3 with a large number of account and about 3500000 record in waveset.log table. Recently we realized through our dba that constantly idm running continue select on a table of logs, causing an excessive use of cpu in oracle and locking then the system. The select is similar to:
SELECT log.id, name, resourceName, accountName, objectType, objectName, action, actionDateTime, actionStatus, subject, sequence from log WHERE ( name='LOG9093A82F9D7001E5:-7BF61D3C:120D0CEFEDD:-254A' )
these select continue for days, without the system makes something special. Is a configuration problem or a know defect solved by workaround or by release upgrade?
My Best regards
Stefano

Answer 1

Ciao Daniela, come sta?
You have a table DAT_MISURA partitioned by TEMPO_ASSOLUTO into 2357 partitions. Then you query based on a different column - N_SEQUENZA_DATO. Unless you specify a specific partition, that means all 2357 partitions must be checked (2357 * index lookup and block read), because Oracle doesn't know the relationship between TEMPO_ASSOLUTO and N_SEQUENZA_DATO (if there is a relationship, that is). And then 2357 result sets have to be merged.
When you query from a specific partition, the difference between first execution and second may mostly be the parsing/optimization phase. There may also be some dynamic sampling going on (see http://www.oracleadvice.com/10g/10g_optimizer.htm) - that could be slow :-)
There's a Jonathan Lewis note about local vs global indexes here: http://www.jlcomp.demon.co.uk/faq/slow_local.html
The second time (16ms) is 4 times faster than the 9i method, it seems. Is that so bad?
Spero che io abbia aiutato!
Tanti saluti
Nigel