Re: Memory keep increasing.

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Sun Nov 15 2009 - 08:59:42 CST

On Sun, 2009-11-15 at 18:14 +0800, Ming wrote:

> Hi Dow,
> Thanks! As you said, I can run the "sync" command when the NAMD job is runing and this can help me decrease the cached memeory. Am I right? Again, how often should I run this command? Many thanks!

ming,

you are on a wild goose chase. the advice about sync is useless.
what sync does, has nothing to do with the consumption of memory by
applications. sync forces outstanding write accesses to be committed
to disk, but that does _not_ free them from the cache. this is standard
unix semantincs, and for certain working on linux kernel, which you are
obviously running.

as i was pointing out before, do you see a steady increase of the memory
usage (you better monitor the namd2 executable(s) rather than the total
system memory)? parallel applications that have internal buffers to
store data that gets passed around (and NAMD qualifies there), usually
only grow their buffer storage, but do not shrink it for performance
reasons (freeing and allocating large chunks of memory costs time). for
NAMD that means that during the equilibration and load balancing phases
of a simulation the memory usage is likely to go up, but it should
level out at some point. if it doesn't, it needs further investigation,
and it would then be imperative that you make your inputs available to
NAMD developers, so they can check it. ...or you do it yourself, e.g.
by running your application under the memory checker tool of valgrind.

cheers,
   axel.
>
> Ming
>
>
>
>
> >>From: Mindspring <Dow.Hurst_at_mindspring.com>
> >>Reply-To:
> >>To: Ming <ebullience_at_emails.bjut.edu.cn>
> >>Subject: Re: namd-l: Memory keep increasing.
> >>Date:Sun, 15 Nov 2009 03:12:43 -0500
> >>
> >>XFS is the file system used on an SGI. It caches data heavily in
> >>RAM. Sync will flush the data to disk.
> >>
> >>Sincerely,
> >>Dow Hurst
> >>
> >>On Nov 14, 2009, at 9:28 PM, "Ming"
> >
> >>wrote:
> >>
> >>> Hi Dow,
> >>> Thanks for your reply. I'm sorry I don't understand what you said.
> >>> I'm very poor on runing a on a SGI station. Could you please tell me
> >>> what XFS mean? I didn'y run sync. Again, I don't know why you ask me
> >>> to run sync. Do you mean I shold tell the system save some data into
> >>> hard drive but not buffer?
> >>>
> >>>
> >>> I ran the "free" command and i didn't find out any decrease in
> >>> cached filesystem data. Any suggestions? Thanks in advance!
> >>>
> >>>
> >>>
> >>>
> >>>> From: Dow_Hurst
> >
> >>>> Reply-To: Dow_Hurst
> >
> >>>> To: Ming
> >
> >>>> Subject: Re: namd-l: Memory keep increasing.
> >>>> Date:Sat, 14 Nov 2009 02:10:51 -0500 (EST)
> >>>>
> >>>> I remember that XFS loves to cache data. Have you run sync on the
> >>>> workstation during the run? And, measured any decrease in cached
> >>>> filesystem data?
> >>>> Dow
> >>>>
> >>>> -----Original Message-----
> >>>>> From: Ming
> >>>
> >>>>> Sent: Nov 13, 2009 12:51 AM
> >>>>> To: akohlmey_at_gmail.com
> >>>>> Cc: namd-l_at_ks.uiuc.edu
> >>>>> Subject: Re: namd-l: Memory keep increasing.
> >>>>>
> >>>>> Hi Axel,
> >>>>> Thanks for your reply. I've been keeping moitering the memory
> >>>>> changes. I found the memory still kept increasing while the job is
> >>>>> runing. Below is the output of the "free -m -s 30". I'm sure
> >>>>> nobody else uses this SGI Altix station. Is this increase in
> >>>>> memory normal? Thanks!
> >>>>>
> >>>>> P.S.By the way, I found the used memory increased by 6G within 12
> >>>>> hours.
> >>>>>
> >>>>> total used free shared buffers
> >>>>> cached
> >>>>> Mem: 61521 11543 49978 0
> >>>>> 0 166
> >>>>> -/+ buffers/cache: 11377 50144
> >>>>> Swap: 9999 0 9999
> >>>>>
> >>>>> total used free shared buffers
> >>>>> cached
> >>>>> Mem: 61521 11553 49968 0
> >>>>> 0 171
> >>>>> -/+ buffers/cache: 11382 50139
> >>>>> Swap: 9999 0 9999
> >>>>>
> >>>>> total used free shared buffers
> >>>>> cached
> >>>>> Mem: 61521 11555 49965 0
> >>>>> 0 171
> >>>>> -/+ buffers/cache: 11384 50137
> >>>>> Swap: 9999 0 9999
> >>>>>
> >>>>> total us
> >>
>

-- 
Dr. Axel Kohlmeyer  akohlmey_at_gmail.com 
Institute for Computational Molecular Science
College of Science and Technology
Temple University, Philadelphia PA, USA.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:30 CST