Hans Buschmann <buschmann@nidsa.net> writes:
> In my opinion, even for Linux the default for parallel_setup_cost is set too low (1000). It should reflect the
sequentialaccess of 1000 pages, which normally is faster from buffer cache on modern hardware.
I tend to agree with you that the default costs for parallel queries
are far too optimistic --- we've seen enough complaints about parallelism
making things slower to make that clear. However, proposing some new
numbers without gathering some hard evidence to support them is not
helpful.
regards, tom lane