linux_AMD64 cmsearch memory requirements
-
cisf
- Mikrocruncher

- Beiträge: 20
- Registriert: 07.03.2010 21:55
linux_AMD64 cmsearch memory requirements
what are the memory requirements for cmsearch ?
in the F.A.Q it says : usually below 100 MB of RAM
but i notice it requires close to 2G ! (http://www.rnaworld.de/rnaworld/result. ... id=1290834)
also i see there are a few "Maximum memory exceeded" errors to other people.
my system starts to swap on such memory requiremnts, and so gets unusable for me, so do i need to check out cmsearch from my crunching options, or it is just an error ?
in the F.A.Q it says : usually below 100 MB of RAM
but i notice it requires close to 2G ! (http://www.rnaworld.de/rnaworld/result. ... id=1290834)
also i see there are a few "Maximum memory exceeded" errors to other people.
my system starts to swap on such memory requiremnts, and so gets unusable for me, so do i need to check out cmsearch from my crunching options, or it is just an error ?
-
yoyo
- Vereinsvorstand

- Beiträge: 8170
- Registriert: 17.12.2002 14:09
- Wohnort: Berlin
Re: linux_AMD64 cmsearch memory requirements
Can you give an reference for this?cisf hat geschrieben: also i see there are a few "Maximum memory exceeded" errors to other people.
I do not see it in the results yet.
yoyo
-
cisf
- Mikrocruncher

- Beiträge: 20
- Registriert: 07.03.2010 21:55
Re: linux_AMD64 cmsearch memory requirements
http://www.rnaworld.de/rnaworld/result. ... id=1290556
http://www.rnaworld.de/rnaworld/result. ... id=1289905 - this computer has about 5 tasks witch errord with memory related issues (to little memory perhaps, but shuldn't the aplicatin test if the computer can handle the jobs ?)
http://www.rnaworld.de/rnaworld/result. ... id=1289905 - this computer has about 5 tasks witch errord with memory related issues (to little memory perhaps, but shuldn't the aplicatin test if the computer can handle the jobs ?)
-
yoyo
- Vereinsvorstand

- Beiträge: 8170
- Registriert: 17.12.2002 14:09
- Wohnort: Berlin
Re: linux_AMD64 cmsearch memory requirements
This one is strange, because it doesn't start the wu, means not even the unzip at the beginning wasn't done. I think this is not an to less ram problem.cisf hat geschrieben: http://www.rnaworld.de/rnaworld/result. ... id=1289905 - this computer has about 5 tasks witch errord with memory related issues (to little memory perhaps, but shuldn't the aplicatin test if the computer can handle the jobs ?)
The other wus aborted because of disc usage. The wu used more disc space than allowed.
I assigned more disc space beginning with wuID=511915.
yoyo
-
Ananas
- WU-Schieber

- Beiträge: 1184
- Registriert: 27.04.2008 18:37
- Wohnort: Nordlichter Köln
Re: linux_AMD64 cmsearch memory requirements
@Yoyo : Some results on this host complain about RAM toocisf hat geschrieben:http://www.rnaworld.de/rnaworld/result. ... id=1290556
http://www.rnaworld.de/rnaworld/result. ... id=1289905 - this computer has about 5 tasks witch errord with memory related issues (to little memory perhaps, but shuldn't the aplicatin test if the computer can handle the jobs ?)
I wonder which setting it is - rsc_memory_bound in the workunit or ram_max_used_busy_pct / vm_max_used_pct in the global preferences.
The Athlon XP might not be allowed to use enough memory for those results - the BOINC default setting for Use at most xx% of memory when computer is in use and Use at most xx% of page file are quite low.
We already had the problem that the WUs went to hosts with insufficient physical RAM when the project has been closed Beta, I haven't seen a report anymore after Yoyo has increased the requirements though.
Especially with results that don't checkpoint (requiring "Leave applications in memory"), the BOINC strategy of pausing results when the free memory gets low is quite destructive. It cannot get any better when all the stuff has to be kept in memory.
vi BOINC/checkin_notes
:1,$s/bug/feature/g
:wq!
Erzaehlen sich Biologen eigentlich Klein-RNA-Witze?
:1,$s/bug/feature/g
:wq!
Erzaehlen sich Biologen eigentlich Klein-RNA-Witze?
-
Michael H.W. Weber
- Vereinsvorstand

- Beiträge: 23174
- Registriert: 07.01.2002 01:00
- Wohnort: Marpurk
Re: linux_AMD64 cmsearch memory requirements
At present, there is no way to exactly determine the RAM requirements except for running the WU - although developments are underway to incorporate a RAM forecast function like the run time forecast we already have available. So it is very useful for us if you report a higher memory requirement than that what we measured as the current maximum with previous WUs (as noted in the FAQ). Please note that these on purpose are indeed very large CMS WUs and the one you have picked ranges amongst the most demanding ones.cisf hat geschrieben:what are the memory requirements for cmsearch ?
Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.


-
yoyo
- Vereinsvorstand

- Beiträge: 8170
- Registriert: 17.12.2002 14:09
- Wohnort: Berlin
Re: linux_AMD64 cmsearch memory requirements
I increased
rsc_memory_bound from 500MB -> 1GB
and
rsc_disk_bound from 100MB 1GB
yoyo
rsc_memory_bound from 500MB -> 1GB
and
rsc_disk_bound from 100MB 1GB
yoyo
-
Michael H.W. Weber
- Vereinsvorstand

- Beiträge: 23174
- Registriert: 07.01.2002 01:00
- Wohnort: Marpurk
Re: linux_AMD64 cmsearch memory requirements
It seems that some people expercience RAM requirements of above 1.5 GB, so I think we need to refine --mxsize and increase the aforementioned boundaries even further for these complex CMS WUs.
Michael.
Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.


-
Ananas
- WU-Schieber

- Beiträge: 1184
- Registriert: 27.04.2008 18:37
- Wohnort: Nordlichter Köln
Re: linux_AMD64 cmsearch memory requirements
The major problem with RAM requirements above 1GB is, that nearly all hosts that have quite a a lot of RAM are set to run several tasks. So even even with 4GB installed, a quadcore might run into a RAM shortage with 4 CMS results running.
p.s.: An important task on the todo list should be to catch the stderr output of cmsearch. As long as stderr always gets lost, it is hard to figure out what caused the program to abort the calculation (if it is a program controlled exit and not a crash).
p.s.: An important task on the todo list should be to catch the stderr output of cmsearch. As long as stderr always gets lost, it is hard to figure out what caused the program to abort the calculation (if it is a program controlled exit and not a crash).
vi BOINC/checkin_notes
:1,$s/bug/feature/g
:wq!
Erzaehlen sich Biologen eigentlich Klein-RNA-Witze?
:1,$s/bug/feature/g
:wq!
Erzaehlen sich Biologen eigentlich Klein-RNA-Witze?