Daniel Morgan <damorgan_at_x.washington.edu> wrote in message news:<1075948868.103447_at_yasure>...
> Database Guy wrote:
>
> > Since RAC is intended to "scale out", can you can answer a question
> > that has been puzzling me?
> >
> > Of its 11 best TPC-H results, all on 9i or 10g, why has Oracle chosen
> > to run only one benchmark under RAC? How come that single (Oracle 10g,
> > 1000GB category) RAC result is easily the worst of all 11 Oracle
> > results?
>
> As has been frequently stated in the DB2 usenet group and elsewhere ...
> the TPC benchmarks favor shared-nothing. Pull the plug on one of those
> shared nothing nodes (hardware failures do happen in the real world) and
> see what happens to performance.
>
> My point being that if I made one change to the benchmark ... say added
> the following ... "one hour into the test pull the plug on one node and
> complete the job" ... shared nothing and federated databases wouldn't
> even be able to compete.
That's true, but for me real performance measurement should be done
while everything is working. According to my experiences If you have
proper enteprise environment outages are almost NULL and during short
period, when you are dealing with some problem, lower performance is
usually acceptable by business users.
On the other hand I apriori distrust *any* benchmarks, because for me
they're just marketing (read lies).
--
Dusan Bolek
Received on Thu Feb 05 2004 - 04:07:33 CST
