Re: Non-linear Performance

On Sat, 08 Jun 2002 00:28:18 -0400
"Doug Fields" <dfields-pg-general(at)pexicom(dot)com> wrote:
> My biggest fear is that the added CPU speed will be sub-linear, and that
> the real speed improvement would come if PG would multi-thread individual
> queries.

I'll confess that I can't understand the need for this feature
(even if one ignores the complexity of implementing it).

On most heavily loaded database servers, isn't the number of concurrent
queries far greater than the number of processors? In which case, what
is the point of allowing a single query to be executed by multiple CPUs
concurrently? If your queries are CPU-bound, the other CPUs are going
to be busy executing other queries at the same time (not to mention that
the cost of migrating processes from one CPU to another and back again
would not help).

I can see the benefit when you're executing a small number of
enormously expensive queries (and those queries are CPU-bound);
but that situation is quite rare, IME.

Cheers,

Neil

--
Neil Conway <neilconway(at)rogers(dot)com>
PGP Key ID: DB3C29FC