Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.tools -> Re: performance drop of >35% from Oracle 7.1 to 8.05
"Sybrand Bakker" <postbus_at_sybrandb.demon.nl> wrote in message news:960222588.944.0.pluto.d4ee154e_at_news.demon.nl...
snip
> IMO, frankly I think VMS is on the way out. I remember seeing remarks on
KPN
> dropping this platform too. This is of course a pity, but was to be
expected
> after DEC was bought by Compaq
Uhoh, I hope you are wearing your flak jacket... That was a pretty dumb
statement considering the board you are on and all the recent activity in
the VMS
world.
You are right about the two process implementation though.
>
> Regards,
>
> Sybrand Bakker
> Oracle DBA
> former VMS user/enthusiast
>
> Dirk Munk <d.munk_at_kpn.com> schreef in berichtnieuws
> 01bfcf04$73b73ab0$ab4815ac_at_HKTGN0002251944...
> > After an upgrade from VMS 6.2 & Oracle 7.1 to VMS 7.2 & Oracle 8.0.5 we
> > noticed a performance drop of > 35 %.
> >
> > At the same time the buffered IO rate went sky high and CPU utilization
> > went up too.
> >
> > After close examination we found that in the old situation a given
program
> > with SQL statements consisted out of one process that communicated with
the
> > database listner.
> >
> > In the new situation the program is made up out of two processes, one
> > process with the same name as the old process, and a second Oracle
process
> > that communicates with the database listener.
> >
> > The communication between both processes is done by means of mailboxes,
and
> > that explains the terrible loss in performance and the high buffered IO
> > rate & CPU utilization.
> >
> > A mailbox in VMS is a device like any other device, and writing to and
from
> > a mailbox is normal IO with all overhead attached to it.
> >
> > Has anyone experienced this problem before (seems Oracle 7.3.4 already
had
> > this problem), and did anyone find a workaround ?
> >
> > Regards,
> >
> > Dirk Munk
> > KPN Telecom
> >
> >
> >
>
>
Received on Mon Jun 05 2000 - 00:00:00 CDT
![]() |
![]() |