If the relations being JOINed are evaluated on the data-nodes, we currently use the nested loop join. Nested loop join is expensive as it involves rescans and rescans need materialization. Hence, enable merge and hash joins for such relations and push down the required operations like Sort required for these joins.
Partially fixed by 5f534780fcd616c0df4092f7b4bc09c5321d8e3f and 0c99b9b71eff391af867977dc0913e55b2020f37.
These two commits enable the use of hash and merge joins but pushing the sorting operation to datanodes is not implemented yet.
Q3
Done
Thanks Suzuki-san for closing this. The feature was implemented long back,
but the request wasn't closed.
On Mon, Jun 1, 2015 at 12:08 PM, Koichi Suzuki koichi-szk@users.sf.net
wrote:
--
Best Wishes,
Ashutosh Bapat
EnterpriseDB Corporation
The Postgres Database Company
Related
Feature Requests:
#68