On Tue, May 4, 2021 at 7:40 PM Hans Buschmann <buschm...@nidsa.net> wrote: > The problem seems that this (probably inherent) performance disadvantage of > windows is not reflected in the cost model.
https://www.postgresql.org/docs/13/runtime-config-query.html#GUC-PARALLEL-SETUP-COST is for that. It might be interesting to do some profiling to see exactly what is slow. Presumably CreateProcess(), but I wonder why exactly. It'd be nice if we had a way to reuse parallel workers, but alas we do not, yet. Or threads.