There are few things to consider:
- you don't need to use distinct on all columns (and therefore sort all columns)
- you should try to sort in memory, better than on-disk
- it seems that the planner doesn't predict the good number of rows
Regards,
Florian
On Mon, Sep 9, 2019 at 12:46 PM Justin Pryzby <pryzby@xxxxxxxxxxxxx> wrote:
On Mon, Sep 09, 2019 at 02:00:01PM +0530, yash mehta wrote:
> We have a query that takes 1min to execute in postgres 10.6 and the same
> executes in 4 sec in Oracle database. The query is doing 'select distinct'.
> If I add a 'group by' clause, performance in postgres improves
> significantly and fetches results in 2 sec (better than oracle). But
> unfortunately, we cannot modify the query. Could you please suggest a way
> to improve performance in Postgres without modifying the query.
Not sure it helps, but I remember this:
https://www.postgresql.org/message-id/CAKJS1f9q0j3BgMUsDbtf9%3DecfVLnqvkYB44MXj0gpVuamcN8Xw%40mail.gmail.com
