Re: Oracle 8i on RS/6000 AIX question.

"Igor Laletin" <ilaletin_at_usa.net> wrote in message news:f9226414.0204292336.42da5598_at_posting.google.com...
> "Howard Rogers" <Howard.Rogers_at_oracle.com> wrote in message
news:<zwjz8.16$707.267_at_news.oracle.com>...
> > Oracle 8i on RS/6000 AIX question.Is the problem one of length of
> > transactions causing blocking, or concurrency of access? If it's
> > long-running transactions, processor speed would help. But I suspect
it's
> > concurrency issues, in which case more processors is likely to be the
better
> > way forward.
>
> I would second this.
>
> > On the other hand, have you implemented a Multi-threaded Server
> > configuration yet or not? It's quite conceivable that a
properly-implemented
> > MTS configuration would make far more effective use of the resources you
> > already posess, and you wouldn't have to upgrade anything.
>
> Correct me if I'm wrong but MTS wouldn't help with cpu-bound system.
> MTS makes a better use of memory, os resources (proc table) but as far
> as cpu is concerned it's an overhead. After all it takes cycles to
> dispatch requests/responces.
>

It does, that's true.

I think I just got suckered by the "multi-threaded" bit of the MTS name, and typed too quickly thinking that since the context switch which occurs as you poll different processors is extremely expensive compared with the switch between different threads in the one process -and therefore MTS would help. I am of course a moron, since MTS (as I posted just yesterday) has nothing multi-threaded about it at all. My apologies.

However, using the multi-threaded argument, it is obvious that our original poster needs to upgrade to Windows NT. (<g>).

I am glad we agree on concurrency <=> more processors however!

;-0
HJR
> Regards,
> Igor
>
> > There really isn't enough information to go on to offer meaningful
advice
> > beyond these basic generalities.
> >
> > Regards
> > HJR
> >
> > (PS. HTML posting is not exactly recommended)
> >
> > ========
> > "Todd Parnell" <TParnell_at_sauder.com> wrote in message
> > news:D577E02F49A4A8498167BA5E41424D4703C5B674_at_expf1...
> > We are using a dual-processor RISC machine and we are CPU bound at
certain
> > times. We are a pretty standard OLTP type of operation.
> > The question has come up, would we be better off doubling the processor
> > speed or doubling the number of processors?
> > As far as Oracle is concerned, the latter is more expensive.
> > I would appreciate hearing from anyone else who has encountered this.
> > Thanks,
> > Todd Parnell
Received on Tue Apr 30 2002 - 14:53:47 CDT