Re: Extra CPU

It's just an old Pentium Pro 200 dual running NT 4.0 SP6 ( HP Vectra Xu ). I was pointing out the ability to tweak the scheduler ( set affinity ) rather than suggesting use of particular hardware.

-Kevin
"Andrey" <aakit_at_softhome.net> wrote in message news:9bp9fb$16go$1_at_josh.sovintel.ru...
> sorry, kevin.brandx_at_tel.gte.com is rejected with User unknown
>
> can you please post you system config/soft
> because I need same dual P,
> but nowhere to ask for optimal config
>
>
> "Kevin Brand" <kevin.brandx_at_tel.gte.com> wrote in message
> news:9b23b9$q55$1_at_news.gte.com...
> >
> > You can't predict which one, no, but you can set affinity once it's
 running
> > so that you confine it to just one processor.
> >
> > Using a dual Pentium NT 4.0 box and MKS Toolkit, I ran the following
 test:
> >
> > In a dos window running sh.exe I start a CPU muncher in a tight loop:
> > while :
> > do
> > :
> > done
> >
> > Now, task manager shows cpu0 averaging 95% and cpu1 averaging 5% ( or
 so )
> > as well as overall cpu utilization at 54% ( or so ).
> >
> > Now, I set affinity so that my sh.exe executes exclusively on cpu0:
> >
> > Now, task manager shows cpu0 absolutely pegged at 100% and cpu1 very
 near
> > idle. Overall cpu utilization seems the same ( maybe just a few points
 less
> > on average ).
> >
> > -Kevin
> >
> >
> > "Jim Kennedy" <kennedy-family_at_home.com> wrote in message
> > news:OzNA6.2357$%o2.113927_at_news1.sttls1.wa.home.com...
> > > No it should just spin the one the procedure is running on. (you
 can't
> > > predict which one)
> > > Jim
> > >
> > > "Ilya Kuzkin" <elliew_at_hotmail.com> wrote in message
> > > news:9av8be$1h4$1_at_news3.cadvision.com...
> > > > Nope. It's a standard loop construction.
> > > > You can try it yourself:
> > > > begin
> > > > for i in 1..200000 loop
> > > > null;
> > > > end loop;
> > > > end;
> > > >
> > > > Spins the single processor fairly well on any NT platform.
> > > >
> > > > And I really do wonder will it spin all CPUs if we have more than
 one.
> > > >
> > > > Sincerely, Ilya.
> > > >
> > > > Jim Kennedy <kennedy-family_at_home.com> wrote in message
> > > > news:TFwA6.700568$U46.21940953_at_news1.sttls1.wa.home.com...
> > > > > Find out why it is using so much processor time. Is it all
 dynamic
 sql?
> > > > > Jim
> > > > >
> > > > > "Ilya Kuzkin" <elliew_at_hotmail.com> wrote in message
> > > > > news:9ati6g$fai$1_at_news3.cadvision.com...
> > > > > > Hi, All.
> > > > > > Environment:
> > > > > > Oracle 8.0.4 (yes, desupported), Dec Alpha NT 4.0 (of course
 desupported
> > > > > > too) :)
> > > > > > Here's the problem - We've got several concurrent processes and
 one
 of
 them
> > > > > > performs heavy PL/SQL code and another executes queries one
 after
 another
> > > > > > (lets call them Process 1 and Process 2 for the further
 reference).
> > > > > > Process 1 is basically spinning the server for the 80% of its
 working
 cycle
> > > > > > with PL/SQL long lasting loop so the performance of Process 2
 falls
> > > > > > drastically (4 times) and we really cannot afford this.
> > > > > > So, I wonder... If we add the second processor to this dinosaur.
 Would
 it
> > > > > > really improve the situation? Wont it happen that that loop will
 occupy
 and
> > > > > > spin both CPUs as it spins the single one?
> > > > > > My concern is - we cannot just take that proc out of the blue,
 we
 have
 to
> > > > > > destroy the standby server machine so it may be costly for us...
 But
 if
 we
> > > > > > will be able to obtain seriuous benefits from this move, please
 let
 me
 know.
> > > > > >
> > > > > > Sincerely, Ilya.
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>
Received on Fri Apr 20 2001 - 10:59:27 CDT