Re: what advantage for using MTS

Connor McDonald <connor_mcdonald_at_yahoo.com> wrote in message news:<3B505088.4E4F_at_yahoo.com>...
> Mark D Powell wrote:
> >
> > "Howard J. Rogers" <howardjr_at_www.com> wrote in message news:<3b4d57d3$1_at_news.iprimus.com.au>...
> > > Think of it this way.
> > >
> > > I fire up SQL Plus. I thereby obtain a dedicated server process, together
> > > with a private bit of memory space (the PGA). I then go for a smoke break.
> > > Oh, then I go for lunch, take in a bit of shopping, have a snooze on a park
> > > bench, and eventually come back to the office for a bit of a natter with the
> > > office floozies. Finally, about 4 hours later, I issue my long-awaited SQL
> > > command (inevitably, 'select * from emp') and do some real work on the
> > > server.
> > >
> > > For X hours, my server process has been sitting doing squat diddly, and
> > > chewing up CPU cycles and (more importantly) swathes of scarce memory to
> > > boot.
> > >
> > > That's exactly what MTS is designed to avoid (if *I* don't stress a shared
> > > server process, someone else likely will). Wasted resources.
> > >
> > > It's also probably the only way you'll get hundreds of concurrent users
> > > connected to your server, unless you have shares in a Taiwanese silicon
> > > foundry.
> > >
> > > I've actually swung in my opinions in the past few months. I've come to the
> > > conclusion that you need a really good reason *Not* to be running MTS, given
> > > that even under an MTS setup, DBAs and people with heavy needs can acquire a
> > > dedicated server connnection when they need to.
> > >
> > > Regards
> > > HJR
> > >
> > >
> > Howard, it used to be that if a session issued a wait for an alert
> > (dbms_alert) that it would cause queueing for all the session assigned
> > to the same MTS server so if your application made use of alerts you
> > needed to avoid using MTS or assign the alert waiters to a dedicated
> > session which may or may not be an easy thing to do. Question: Do you
> > know if this is still true?
> >
> > -- Mark D Powell --
>
> I can't comment on the specific alert question, but I'm still pretty
> sure that any session deemed as "active" (eg dbms_lock.sleep(n)) will
> grab and not release a shared server...
>
> hth
> connor

Thanks Connor, that is good to know.

• Mark D Powell --