On Tue, Dec 22, 2015 at 2:16 AM, Paul Ramsey <pramsey@cleverelephant.ca> wrote:
> Shouldn’t parallel aggregate come into play regardless of scan selectivity?
> I know in PostGIS land there’s a lot of stuff like:
>
> SELECT ST_Union(geom) FROM t GROUP BY areacode;
>
> Basically, in the BI case, there’s often no filter at all. Hoping that’s
> considered a prime case for parallel agg :)
Yes, the latest patch attached in the thread addresses this issue.
But it still lacks of proper cost calculation and comparison with
original aggregate cost.
The parallel aggregate selects only when the number of groups
should be at least less than 1/4 of rows that are getting selected.
Otherwise, doing aggregation two times for more number of
records leads to performance drop compared to original aggregate.
Regards,
Hari Babu
Fujitsu Australia