AW: Low CPU usage with NAMD running on linux cluster

From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Mon Jul 30 2012 - 11:01:37 CDT

Hi all,

for me this sounds more like a problem of the node interconnect. BTW as long as the scaling is linear, you shouldn't give too much on low CPU utilization. If the scaling is bad, that means the speedups are not the same as the used nodes, you should more worry.

First of all we need to know what interconnect you have. You already told Ethernet, but there are 10/100/1000/10000 MBit/s versions outside there so please tell what it is.

You can check that by doing lshw on the nodes and look for Ethernet adapters, post the name of the devices here.
Also the name or model of the switch is of course important.

Let us know

Norman Geist.

> -----Ursprüngliche Nachricht-----
> Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im
> Auftrag von amin_at_imtech.res.in
> Gesendet: Sonntag, 29. Juli 2012 07:10
> An: Rajan Vatassery
> Cc: namd-l_at_ks.uiuc.edu
> Betreff: Re: namd-l: Low CPU usage with NAMD running on linux cluster
>
> What I meant was that I am distributing 24 processes over 4 nodes
> having 8
> processors each.Therefore each node will have 2 processors free. I have
> not
> completed any run till now so I cant give the "TIMING INFORMATION"
> because the
> simulation is running extremely slow. My 2 ns simulation is running for
> more
> than 2 days now.I will post that information as soon as the run is
> completed.
> Thanks.
>
> Amin.
>
>
>
>
> > Amin,
> > Do you really mean that you have requested 4 nodes x 8 processors
> each
> > = 32 processors? I'm wondering because you said you have only 24
> > processors. Also, Branko is not asking about output frequencies, but
> > rather the details of the output that your simulation is giving. For
> > example, the "TIMING" information from the log file is indicative of
> the
> > seconds/step that your simulation is running at.
> >
> > rajan
> >
> > On Sat, 2012-07-28 at 20:40 +0530, amin_at_imtech.res.in wrote:
> >> Thanks. I have read the link. I have a PBC system with 20,000 atoms
> with time
> >> step=1, dcdFreq=500, outputEnergies=1000. I am trying to run NPT
> simulation
> >> using 24 processors on 4 nodes having 8 processors each.
> >>
> >> Regards.
> >> Amin.
> >>
> >>
> >> > Amin,
> >> >
> >> > Provide more data about the size of your system, output data, and
> see:
> >> >
> >> > http://www.ks.uiuc.edu/Research/namd/wiki/?NamdPerformanceTuning
> >> >
> >> > Branko
> >> >
> >> > On 7/28/2012 8:33 AM, amin_at_imtech.res.in wrote:
> >> >> Dear all,
> >> >> I am trying to run NAMD on a linux cluster. I am using NAMD
> >> Linux-x86_64
> >> >> (64-bit Intel/AMD with ethernet). While i am able to run the
> program on the
> >> >> nodes listed in the nodelist file but I find that all the
> processes are
> >> >> running at only 8-12 % CPU usage. Can someone please guide me?
> >> >>
> >> >> Regards.
> >> >> Amin.
> >> >>
> >> >>
> ______________________________________________________________________
> >> >> सूक्ष्मजीव प्रौद्योगिकी संस्थान (वैज्ञानिक
> औद्योगिक अनुसंधान परिषद)
> >> >> Institute of Microbial Technology (A CONSTITUENT ESTABLISHMENT OF
> CSIR)
> >> >> स%u09
>
> ______________________________________________________________________
> सूक्ष्मजीव प्रौद्योगिकी संस्थान (वैज्ञानिक औद्योगिक
> अनुसंधान परिषद)
> Institute of Microbial Technology (A CONSTITUENT ESTABLISHMENT OF CSIR)
> सैक्टर 39 ए, चण्डीगढ़ / Sector 39-A, Chandigarh
> पिन कोड/PIN CODE :160036
> दूरभाष/EPABX :0172 6665 201-202

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:22:19 CST