Page faults

Everything about the project RNA World
Nachricht
Autor
adrianxw

Page faults

#1 Ungelesener Beitrag von adrianxw » 18.08.2012 09:35

I was curious as to why the disk in this machine, (Win XP 32bit), was thrashing and found that cmsearch S was page faulting like crazy. It has been running 6 minutes so far and has memory usage of 439,832k and 24,413,287 page faults.

Benutzeravatar
Michael H.W. Weber
Vereinsvorstand
Vereinsvorstand
Beiträge: 23174
Registriert: 07.01.2002 01:00
Wohnort: Marpurk

Re: Page faults

#2 Ungelesener Beitrag von Michael H.W. Weber » 18.08.2012 09:37

Maybe check your memory modules?

Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.

Bild Bild

adrianxw

Re: Page faults

#3 Ungelesener Beitrag von adrianxw » 18.08.2012 09:40

I can, but have no suspicion in that direction, nothing else is having any issues.

adrianxw

Re: Page faults

#4 Ungelesener Beitrag von adrianxw » 19.08.2012 16:42

The work unit has finished without issue. High memory requirement and still all that paging...

ChristianB
Admin
Admin
Beiträge: 1920
Registriert: 23.02.2010 22:12

Re: Page faults

#5 Ungelesener Beitrag von ChristianB » 19.08.2012 17:25

page faults just means your machine is paging like crazy. RNAWorld can be very memory consuming and it seems you didn't have enough physical RAM available so the system used your hard-disk page file. If this is happening more often you should upgrade your RAM on this machine if you want to speed up the RNAWorld apps.

adrianxw

Re: Page faults

#6 Ungelesener Beitrag von adrianxw » 21.08.2012 06:16

I have been a professional software engineer for 35 years, I KNOW what a page fault is, believe me! The machine is little more than a BOINC cruncher, (has been so for 5 - 6 years), and I've not noticed that level of thrashing before. It has 4GB of RAM, that is not going to increase. If I see that happening again, I'll pull RNA off there, it crunches everything else okay as it is.

Maybe increase the working set size, but that, of course, puts limits on the types of machine that can run the app...

ChristianB
Admin
Admin
Beiträge: 1920
Registriert: 23.02.2010 22:12

Re: Page faults

#7 Ungelesener Beitrag von ChristianB » 21.08.2012 20:37

sorry, it was not meant as an insult. I just wanted to say that the cmsearch app can be very memory consuming for some data and exceed the available ram.

adrianxw

Re: Page faults

#8 Ungelesener Beitrag von adrianxw » 22.08.2012 07:37

No insult taken, no apology necessary. :wink:

Benutzeravatar
Michael H.W. Weber
Vereinsvorstand
Vereinsvorstand
Beiträge: 23174
Registriert: 07.01.2002 01:00
Wohnort: Marpurk

Re: Page faults

#9 Ungelesener Beitrag von Michael H.W. Weber » 22.08.2012 08:24

Well, the only thing I can tell you is that you are the first and only who reported this problem to us. It would help, if you could let us know how many cores your 32 bit XP machine has, which RNA World WU was giving the problem and, importantly, what else was running on the remaining cores. :D

In case of RNA World XXL WUs, a single WU can sometimes require several GB on its own. If you are running other XXL RNA World WUs in parallel on the (presumed) other cores (or other RAM-consuming tasks), with only 4 GB of RAM you might very well run into memory issues, especially on a 32 bit system.
The most simply way to avoid this is to disable the XXL WUs in the RNA World project settings.

Note that BOINC appears incapable of handling proper task/memory assignment on multicore machines. I have experienced system halts on an 8 GB XP Pro 64 bit quadcore AMD 955BE machine when 4 massive RNA World XXL WUs with runtimes of 1000 hrs were running in parallel.

Good news is that in a future version we will finally have RNA World occupy all cores for a single XXL WU task. This would also have been possible with the current version - actually since January 2010 - but for one there was a problem again with BOINC not supporting this and finally a WU validation problem was there which now has been fixed.

Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.

Bild Bild

adrianxw

Re: Page faults

#10 Ungelesener Beitrag von adrianxw » 22.08.2012 12:38

I may have been the only to report it because I was the only one who noticed it! I would not regard it really as a problem that an end user would see, because of my professional background, I notice things like disk drives thrashing.

The machine is a quad core, an Intel Q6600, at standard clocking. The wu that was causing the "issue" is this one, (I can see from the reported date that it was so). What was running on the other cores is more difficult, almost certainly a couple of SIMAP wu's as that has quite a high priority on this machine, but these are tiny, just 10-12MB. The other core could have been doing almost anything from my BOINC portfolio, which is quite large. Climate Prediction has a fairly large memory footprint at times, so does NumberFields but neither would have been using as much as 1GB. Most of the rest use much less.

The wu is not an XXL but an S.

Zurück zu „RNA World Discussions (english)“