Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Performance impact of inactive sessions

Re: Performance impact of inactive sessions

From: EdStevens <quetico_man_at_yahoo.com>
Date: 21 Jan 2007 20:00:14 -0800
Message-ID: <1169438414.792861.27200@a75g2000cwd.googlegroups.com>

Martin T. wrote:
> EdStevens wrote:
> > Martin T. wrote:
> > > EdStevens wrote:
> > > > sybrandb wrote:
> > > > > On Jan 19, 10:00 am, "Jack" <n..._at_INVALIDmail.com> wrote:
> > > > > > Shortly: Nope
> > > > > >
> > > > > > "Martin T." <bilbothebagginsb..._at_freenet.de> wrote in messagenews:1169195643.040984.149850_at_v45g2000cwv.googlegroups.com...
> > > > > >
> > > > > >
> > > > > >
> > > > > > > Greetings!
> > > > > >
> > > > > > > Please bear with me if I'm being somewhat unspecific here, but I yet
> > > > > > > don't know enough to get more specific:
> > > > > >
> > > > > > > I will make it short for this initial question: Would you expect a
> > > > > > > problematic performance impact on a Oracle 9i2 database by lots (say 30
> > > > > > > on a decent modern workstation PC) of INACTIVE sessions versus a system
> > > > > > > where the number of active processing (say 4-5 active sessions) is the
> > > > > > > same without these inactive sessions?
> > > > > >
> > > > > > > thank you,
> > > > > > > br,
> > > > > > > Martin- Hide quoted text -- Show quoted text -
> > > > >
> > > > >
> > > > > Disagree.
> > > > > Inactive sessions don't release their memory.
> > > > >
> > > > > --
> > > > > Sybrand Bakker
> > > > > Senior Oracle DBA
> > > >
> > > > True, but they also don't consume any CPU or I/O. The only way I could
> > > > see an inactive session impacting performance would be if the memory
> > > > consumption forced additional OS paging. And even at that, once the
> > > > memory for the inactive session is paged out, it should pretty much be
> > > > out of the picture until such time as it 'goes active' and has to be
> > > > paged in.
> > > >
> > > > I guess it comes back to .. if you have a performance issue, identify
> > > > the actual bottleneck and address the issue.
> > >
> > >
> > > Thanks all for the replies.
> > >
> > > And yes, this post was part of the search for the bottleneck. (Because
> > > we have not the tiniest clue at the moment as to where that might be
> > > :-/ )
> > >
> > > br,
> > > Martin
> >
> > So, pull a 10046 trace or a very tightly focused statspack and see what
> > your biggest wait events are.

>

> Hee, hee. Just found about statspack yesterday. :-|
> Did a report and I'm currently trying to figure it out.
> (Asked a colleague here if he knew how to read such a report, but he's
> worked with Oracle for the last 5 years and never user such a thing
> (didn't know what it is) ... there you go.

Oh, my! Worked with Oracle five years and didn't know what statspack was? Oh, my!

Ok, you *must* read up on how to install statspack and generate a report. Search MetaLink.

After you have a statspack report in hand, log on to www.oraperf.com and feed the report into the statspack analyzer there. You will get back a nicely formated analysis of your raw statspack data, along with ranked suggestions. Be sure you consider the time frame you are measuring vis a viz the time frame the problem occurrs. Studying the output of the oraperf analysis will be very enlightening, but will probably still leave you with lots of questions. But at least at that point you will be able to come back here with more focused questions and more detail with which to provide some context.

>Me, I never wondered why we

> had problems with the DB ... just trying (and up to now failing 1/2 the
> time) to fix things ...)
>
> cheers,
> Martin
Received on Sun Jan 21 2007 - 22:00:14 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US