On Tue, Nov 24, 2015 at 8:59 AM, Robert Haas <robertmh...@gmail.com> wrote: > > > One idea about parallel sort is that perhaps if multiple workers feed > data into the sort, they can each just sort what they have and then > merge the results.
Sounds like a good approach for parallel sorting, however small extension to it that could avoid merging the final results is that workers allocated for sort will perform range-based sorting. A simple example to sort integers from 1-100 will be, worker-1 will be responsible for sorting any integer between 1-50 and worker-2 will be responsible for sorting integers from 51-100 and then master backend just needs to ensure that it first returns the tuples from worker-1 and then from worker-2. I think it has some similarity to your idea-5 (use of repartition), but not exactly same. With Regards, Amit Kapila. EnterpriseDB: http://www.enterprisedb.com