Re: Speed up transaction completion faster after many relations are accessed in a transaction

On Tue, 19 Feb 2019 at 12:56, Andres Freund <andres(at)anarazel(dot)de> wrote:
> Isn't a large portion of benefits in this patch going to be mooted by
> the locking improvements discussed in the other threads? I.e. there's
> hopefully not going to be a ton of cases with low overhead where we
> acquire a lot of locks and release them very soon after. Sure, for DDL
> etc we will, but I can't see this mattering from a performance POV?

I think this patch was born from Amit's partition planner improvement
patch. If not that one, which other threads did you have in mind?

A problem exists where, if using a PREPAREd statement to plan a query
to a partitioned table containing many partitions that a generic plan
will never be favoured over a custom plan since the generic plan might
not be able to prune partitions like the custom plan can. The actual
problem is around that we do need to at some point generate a generic
plan in order to know it's more costly and that requires locking
possibly every partition. When plan_cache_mode = auto, this is done
on the 6th execution of the statement. After Amit's partition planner
changes [1], the custom plan will only lock partitions that are not
pruned, so the 6th execution of the statement bloats the local lock
table.

[1] https://commitfest.postgresql.org/22/1778/

--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services