On Tue, Nov 30, 2004 at 08:34:20AM +0100, Michael Paesold wrote:
> Mark Wong wrote:
>
>
> >I have some initial results using 8.0beta5 with our OLTP workload.
> > Off the bat I see about a 23% improvement in overall throughput. The
> > most significant thing I've noticed was in the oprofile report where
> > FunctionCall2 and hash_seq_search have moved down the profile a bit.
> >
> > Also, I have libc with symbols now so we can see what it's doing with
> > in the oprofile output.
> >
> > 8.0beta5 results:
> > http://www.osdl.org/projects/dbt2dev/results/dev4-010/199/
> > throughput: 4076.97
> >
> > 8.0beta4 results:
> > http://www.osdl.org/projects/dbt2dev/results/dev4-010/191/
> > throughput: 3323.07
>
>
> Is it possible that there are some other differences effecting the result?
> At "Table Blocks Read" there is a "history" table in #191, but it does not
> show up in #199. Just a thought,... but can you explain, Mark?
>
> Best Regards,
> Michael
Ah, the chart differences are due to a faulty script. That should be
cleared up and updated shortly. As for other differences to the beta4
test, I ran the beta5 test for 30 minutes as opposed to 60,
random_page_cost=2 instead of 4, and I have archving disabled. I've
generally found the latter two settings to have minimal effects but I
can always reverify.
Mark
