Re: [K12OSN] freezing - 100% cpu utilization

jam@McQuil.com
Wed, 2 Oct 2002 19:37:57 -0400 (EDT)


Mike,

I think that IF you have a runaway process, then verynice would
help quite a bit.

But, it doesn't sound like that is the case here.  It sounds
like gcompris is just really hogging alot of cpu time.
Perhaps it was just never written to be multiuser friendly.

I think that dual cpus will definately help.  If you have
a cpu intensive app, it can hog an entire cpu.  if that happens
to be the only cpu you have, then everybody suffers.

At least with dual cpus, you stand a chance of getting some
decent response time when one app begins hogging a cpu.

But, if 2 apps start hogging the cpu, then you may be right
back where you started.

Jim.



On Wed, 2 Oct 2002, msc wrote:

> Hi Jim,
> 
> Gcompris is sort of all inclusive - it has some math, geography, play stuff
> in it.  I know portions of it are cpu intensive because if the kids are in
> different areas the server responds differently - read "better".  I don't
> know what it's written in but I think I'll look to see if there's an update
> (or try something else altogether - it's just a cool program).  I'll have to
> talk to my boss to see about dual cpu's.  What do you think of Todd O'Bryans
> post about VeryNice?  Am I go about solving it a bit backwards with that?
> Do you think the load avg. would get dropped significantly w/ dual's?
> 
> Thanks,
> Mike
> 
> ----- Original Message -----
> From: 
> To: 
> Sent: Tuesday, October 01, 2002 6:05 PM
> Subject: Re: [K12OSN] freezing - 100% cpu utilization
> 
> 
> > Mike,
> >
> > I don't know anything about gcompris.  What does it do ?
> > What is gcompris written in ?  Maybe an interpreted language
> > that is interpreting all the time.
> >
> > Are there options in gcompris to turn off certain things, similar
> > to the way Mozilla allows you to turn off gif animations ?
> >
> > A load avg of 9.29 seems REALLY high.
> >
> > I've got a dual P3-650 server running 120 users, and about 70
> > of those users are running Netscape and OpenOffice.  The load
> > avg rarely goes above 1.0.
> >
> > As for single p4 / dual p3, I ALWAYS go for the dual processors.
> > In a multi user system, dual cpus, even running at half the
> > clock speed, seems to handle the load better.  At any given time,
> > one user may hog the cpu for a short time, meanwhile, other users
> > don't notice so much.
> >
> > In your case, it just sounds like gcompris is a very cpu intensive
> > application.  To fix it, you either need to figure out a way
> > to reduce how much cpu the app needs, or increase the horsepower
> > of the server.
> >
> > Jim McQuillan
> > jam@Ltsp.org
> >
> >
> > On Tue, 1 Oct 2002, msc wrote:
> >
> > > Hi Jim,
> > >
> > > The only thing running besides the services themselves is gcompris.
> Each
> > > user has an instance next to his login name.  Load avg shows something
> like
> > > 9.29  7.77  4.23.  I'm not currently running an internet browser... soon
> > > though if I can get the main bugs worked out.  There are 30 workstations
> in
> > > the lab - today 16 students were using the server - the biggest class is
> 22
> > > or 24.  Generally all the systems are booted but not all are populated.
> As
> > > far as runaway processes, I didn't notice any with top.  Again, If I had
> the
> > > students stop everything, my idle time would come back to normal.  I can
> > > maybe propose dual cpu's but it has to fix it.  I thought I read here
> that a
> > > single fast p4 is as good as two pIII's.  Is that not right?
> > >
> > > Thanks,
> > > Mike
> > >
> > > ----- Original Message -----
> > > From: 
> > > To: 
> > > Sent: Tuesday, October 01, 2002 4:28 PM
> > > Subject: Re: [K12OSN] freezing - 100% cpu utilization
> > >
> > >
> > > > Mike,
> > > >
> > > > Besides Icewm, what programs are running ?
> > > >
> > > > What does 'top' show ?
> > > >
> > > > How about the load avg (uptime, w and top will show this).
> > > >
> > > > I've seen silly things like animated gifs on webpages chew
> > > > up gobs of CPU time.  For that, you can set Mozilla to
> > > > either not run the animations, or just cycle the animation
> > > > 1 time, when you first get to a new web page.
> > > >
> > > > How many workstations are connected ?
> > > >
> > > > My hunch is you have a runaway process, like Mozilla that
> > > > is consuming all available cpu.  This is where a dual cpu
> > > > machine really shines.
> > > >
> > > > Jim McQuillan
> > > > jam@Ltsp.org
> > > >
> > > >
> > > >
> > > >
> > > > On Tue, 1 Oct 2002, Julius Szelagiewicz wrote:
> > > >
> > > > > Mike,
> > > > > you've run out of processing power. add more processors or if you
> > > > > can not, replace the server with multiprocessor box. you might
> expell
> > > some
> > > > > kids, but that would not be nice, ;-)
> > > > > julius
> > > > >
> > > > > On Tue, 1 Oct 2002, msc wrote:
> > > > > > I've posted here before regarding this.  My server is now a single
> > > 2.4hz
> > > > > > w/ 2GB rambus memory (old config was p4 1.7 and 768mb then 15.gb
> > > > > > memory).  It has an intel 10/100 nic and a 3com 10/100nic.  It
> does
> > > have
> > > > > > a single ata100 ide drive (bad I know but I don't think it is
> causing
> > > my
> > > > > > issue).  I have added the max-files per your recommendation (first
> > > 32000
> > > > > > now 64000) to the rc.local file.  Top and free show plenty of free
> > > > > > memory and no paging.  I think my memory used to be an issue but
> is
> > > now
> > > > > > resolved.  I am running icewm on all the stations.  The hard drive
> > > light
> > > > > > shows activity but it is not pegged or on all the time (i.e on a
> > > windows
> > > > > > box running defrag - thats on all the time).  The older kids logon
> cpu
> > > > > > %60 used.  They'll start working and it will be approx- 90% used
> by
> > > > > > users and 8% used by the system w/ 0.0% idle - here comes the
> > > freezing.
> > > > > > If the problems escalate the whole class can stop for 5 seconds
> and
> > > the
> > > > > > cpu recovers to 90% idle.  The younger kids idle time will dip to
> 0%
> > > but
> > > > > > recovers back to 20 - 30 % idle - I think because they work
> slower.
> > > > > > Help please. I'm going to post some other questions for this lab
> > > because
> > > > > > I'm optimistic this will be resolved :-).
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > _______________________________________________
> > > > > K12OSN mailing list
> > > > > K12OSN@redhat.com
> > > > > https://listman.redhat.com/mailman/listinfo/k12osn
> > > > > For more info see 
> > > > >
> > > >
> > > > --
> > > >
> > > >
> > > >
> > > > _______________________________________________
> > > > K12OSN mailing list
> > > > K12OSN@redhat.com
> > > > https://listman.redhat.com/mailman/listinfo/k12osn
> > > > For more info see 
> > >
> > >
> > >
> > > _______________________________________________
> > > K12OSN mailing list
> > > K12OSN@redhat.com
> > > https://listman.redhat.com/mailman/listinfo/k12osn
> > > For more info see 
> > >
> >
> > --
> >
> >
> >
> > _______________________________________________
> > K12OSN mailing list
> > K12OSN@redhat.com
> > https://listman.redhat.com/mailman/listinfo/k12osn
> > For more info see 
> 
> 
> 
> _______________________________________________
> K12OSN mailing list
> K12OSN@redhat.com
> https://listman.redhat.com/mailman/listinfo/k12osn
> For more info see 
> 

-- 



_______________________________________________
K12OSN mailing list
K12OSN@redhat.com
https://listman.redhat.com/mailman/listinfo/k12osn
For more info see