On Mon, Aug 2, 2010 at 6:07 PM, Greg Smith <greg@2ndquadrant.com> wrote:
> Josh Berkus wrote:
>>
>> That doesn't make much sense unless there's some special advantage to a
>> 4K blocksize with the hardware itself.
>
> Given that pgbench is always doing tiny updates to blocks, I wouldn't be
> surprised if switching to smaller blocks helps it in a lot of situations if
> one went looking for them. Also, as you point out, pgbench runtime varies
> around wildly enough that 10% would need more investigation to really prove
> that means something. But I think Yeb has done plenty of investigation into
> the most interesting part here, the durability claims.
Running the tests for longer helps a lot on reducing the noisy
results. Also letting them runs longer means that the background
writer and autovacuum start getting involved, so the test becomes
somewhat more realistic.