Hi,

since recently one of my BOINC clients is starting all work units, which results in all of them being in memory at once. So there's always about 6 to 7 processes waiting in background and silently eating up ~300 MB of precious memory each. As a consequence my computer swapping stuff all the time, making it a pain to work with.

Does anyone know this problem or has seen something similar? I'd greatly appreciate any hint. :-)

Ah, by the way. Here is a screenshot showing BoincView with the 3 computers I run Boinc on:
http://www.blamestar.de/screen01.jpg
LAEPTOP, the one in question, is running Windows XP while the other two run a copy of Ubuntu.

Thanks for your efforts,
Fritz Webering

What version of BOINC?

---

Kathryn :o)

Usually once a task has been suspened for some time it will be swapped out to the disk leaving your main memory free for other things.

How much main memory does your laptop have and what are your memory preferences set to? One thing that can cause several tasks to be running at the same time is when they run out of usable memory, BOINC will suspend them and start another until it runs out of memory.

---

BOINC WIKI

BOINCing since 2002/12/8

Thanks for the fast replies.

Quote from my message log:
"Starting BOINC client version 5.10.28 for windows_intelx86"

Keck_Komputers:
True, but swapping several hundret MB to a slow laptop hard drive can take some time. I recently upgraded to 2 GB of memory. The theoretical maximum for virtual memory is 5 GB (2 GB RAM + 3 GB swap file). I practice however I never got near that.

Anyways, isn't that paradoxical? When it has no more memory, why should it start a new process that eats up even more memory? And even if it would: it's quite unlikely that BOINC even touches the boundary of the RAM. Usually there's about 1.6 to 1.8 GB used at most.

Best regards,
a slightly confused IT student

If you're having that many problems with memory, try disabling "keep suspended work in memory".

---

This often happens when a project with variable deadline meets a large cache on a host with multiple CPUs.

The simulation does not simulate what would happen if the most critical workunit would be started after finishing the one that is currently running, it simulates only the worst case scenario.

This means that switching to EDF never tries to finish the active WUs and then start panic mode, it always goes into panic mode immediately.


edit : In this specific situation it might help to reduce the target runtime of RALPH on the venue for the laptop, reduce the cache size and contact the project once. Leave it like that until you're back to round robin mode. When round robin is back, increase the target runtime in small steps until it's back to the value that you prefer.

Hey, it looks like the problem is gone. It seems that Keck_Komputers was right. The BOINC client was configured to use no more that 20% of the available memory while the computer is in use. On a 2GB machine this means 400MB max. for BOINC and obviously there is no way to fit two 300MB blocks into 400MB. So I increased it to allow BOINC to use 35% (about 700 MB) of the memory regardless of the computer being used. That seems to have solved the problem. At the moment there's only two WUs running and hopfully it will stay that way.

So thanks again for your efforts :)
Fritz