On Tue, 2009-07-14 at 11:23 +0300, Heikki Linnakangas wrote:
> > The single SQL example mentioned already has at least two mechanisms for
> > improving performance of that type of query. We probably don't need
> > another, or at least we need a good analysis of why.
>
> Well, another class of queries where index-only scans are beneficial is
> when you fetch a range of rows from index, where the heap fetches result
> in a lot of random I/O.
Which we just optimised, no? I see it could be even better, but do we
really need that class of query to be optimized again? (Possibly...)
I'm not doubting your ability to solve every problem we face, just
advising that we crunch a few numbers before we spend months on
implementing all of this. It would be useful to have projected gains and
a useful test case ahead of time, so we can measure the potential as
early as possible. Yes, you don't need to convince me before you
proceed, but if you can at least share the thoughts that have convinced
you that would help.
-- Simon Riggs www.2ndQuadrant.comPostgreSQL Training, Services and Support