On Feb 21, 2008, at 12:28 PM, Guillaume Cottenceau wrote:
> I have made a comparison restoring a production dump with default
> and large maintenance_work_mem. The speedup improvement here is
> only of 5% (12'30 => 11'50).
At one point I was evaluating several server vendors and did a bunch
of DB restores. The one thing that gave me the biggest benefit was to
bump the number of checkpoint segments to a high number, like 128 or
256. Everything else was mostly minor increases in speed.