Thread: Pre-calculate hash join

When joining two large tables [common in warehousing], a hash join is commonly selected. Calculating hash values for
themerge phase is CPU intensive. Is there any way to pre-calculate value hashes to save that time? Would it even grant
anyperformance to skip the build phase of the hash join? 

-Warren

warren@warrenandrachel.com writes:

> When joining two large tables [common in warehousing], a hash join is
> commonly selected. Calculating hash values for the merge phase is CPU
> intensive. Is there any way to pre-calculate value hashes to save that
> time? Would it even grant any performance to skip the build phase of
> the hash join?

Maybe maintaining a materialized view then using it in the queries would
be sufficient to solve your problem?

--
dim

Dimitri Fontaine skrev 2010-01-12 12.01:
> warren@warrenandrachel.com writes:
>
>> When joining two large tables [common in warehousing], a hash join is
>> commonly selected. Calculating hash values for the merge phase is CPU
>> intensive. Is there any way to pre-calculate value hashes to save that
>> time? Would it even grant any performance to skip the build phase of
>> the hash join?
>
> Maybe maintaining a materialized view then using it in the queries would
> be sufficient to solve your problem?
>

Depending on your case creating a temporary table in the beginning of
the session might save you a lot of time if multiple similar queries is
done.

Regards,
roppert