Re: Parallel query

Michiel Brunt <mbrunt_at_inergy.nl> wrote:
>We use Oracle 8i for a data warehouse. There is a large (>20 mln) fact table
>and smaller dimension tables (max 100.000).
>Parallel processing should be useful, but what tables should be altered for
>parallel query?

Do you mean partitioning for parallel update and delete? Selects are parallelised just by having parallel server enabled, and putting a parallel clause in the select.

Think about exactly what you're trying to achieve by partitioning. If it's just performance through parallel query, then just hash partitioning may be sufficient.

If you're going to the effort of partitioning though, it is worth looking for "natural" boundaries to range partition on. This could allow dramatic efficiencies through, for example, archiving or deleting old data with a partition exchange, or a partition drop.

-- 
Andrew Mobbs - http://www.chiark.greenend.org.uk/~andrewm/