Re: Rapidly decaying performance repopulating a large table

"David Wilson" <david.t.wilson@gmail.com> writes:
> My guess at this point is that I'm just running into index update
> times and checkpoint IO. The only thing that still seems strange is
> the highly variable nature of the COPY times- anywhere from <1.0
> seconds to >20 seconds, with an average probably around 8ish. I can
> live with that, but I'm still open to any other suggestions anyone
> has!

What have you got shared_buffers set to?  If it's not enough to cover
the working set for your indexes, that might be the (other) problem.

            regards, tom lane