Re: Stats on Huge Partitioned Table and ORA-01555

> I think you can use historical table ( copy of your current table ) to save no needed data, and the current table contains only need data for daily
> processing, reporting...- for example data for the last year …

I think this is the reason why they have designed/partitioned the table - each partition is only round about 6 GB (= 3072 GB / 503) and maybe some legal guidelines have not been reached yet (e.g. 10 years for financial data). They just need to introduce incremental statistics for such a design (i assume on the given data).

_at_ Michael
> I don't know what incremental statistics means so I will look that up after I send this email

Doug Bourns presentation about (incremental) statistics on partitioned objects is one of the best sources to start with i guess. URL: http://de.slideshare.net/dougburns/statistics-on-partitioned-objects

… and if you want the details about it. You can check out Amit Poddar's paper - starting with chapter "4.2 Partition statistics and Global statistics": https://jonathanlewis.files.wordpress.com/2011/12/one-pass-distinct-sampling.pdf

Best Regards
Stefan Koehler

Freelance Oracle performance consultant and researcher Homepage: http://www.soocs.de
Twitter: _at_OracleSK

> Cherif Ben Henda <cherif.benhenda_at_gmail.com> hat am 27. Mai 2015 um 11:33 geschrieben:
>
> Hi Michael,
>
> Why having "Table has 7.1 billion+ rows, 503 partitions, 3TB" ?
> I think you can use historical table ( copy of your current table ) to save no needed data, and the current table contains only need data for daily
> processing, reporting...- for example data for the last year ...
>
> Otherwise ,You can have several performances issues, because if the gathering stats job fails to update stats on others table with 10% change on
> time (night window), we will finish with tables with no up to date statistics.
>
> As said by Stefan, you can think to SET_TABLE_PREFS.
>
> Thanks,
> Cherif

--
http://www.freelists.org/webpage/oracle-l