Re: sane number of the table partitions in DWH
Date: Wed, 1 Apr 2009 08:39:56 -0500
>> Partioning pruning is obviously not for free
Of course. I didn't mean that partition pruning is free and my statement was in the scope of hard parsing only, not execution. Let me clarify this further:
Let's say that partition count is higher and partition pruning is done at parse time (meaning use of literal variables on partition keys). Then during parse time Optimizer must choose the correct partition comparing boundaries of partitions. My point was that, as long as, your buffer_cache and shared_pool are decently sized,selection of a partition at parse step is not costly. Further to the point, rowcache aids hugely for this.
I guess, I should have used the term 'partition selection at parse time'. Of course, I didn't consider 'partition pruning' at execution time with the details provided.
Actual, partition pruning at execution step is a completely different matter :-)
Reposting due to overquoting..
> >> Partioning pruning is obviously not for free