Hi Everyone,

I just noticed this a few weeks ago. I have Seti and Rosetta running under Boinc 5.10.30. When they are idle the processes are remaining in memory. Rosetta is taking up almost 100MB and Seti is taking up about 42MB. I have checked the prefs for both projects and the box for "remain in memory" is unchecked. Also I have clicked the clear button to make sure the projects are not using local prefs which shouldn't matter since the check box is clear there also. How do I get the projects to unload when the PC is being used?

Kage_

Just try a simple thing: Open BOINC manager, Advanced view, Projects tab, select either Seti or Rosetta and click Update.

Else, post the contents of your global_prefs.xml file.

---

I cant see it unloading every-time a system is being used intermittently e.g....could as well exit the client or set the schedule to wait a long time before resuming. Maybe it should be pushed off into swap file?

I don't think it's possible to "push it off to swap". And anyway, I think it would be counter-productive to do so. It would cause lag while the OS saves all that to disk. It's better to just leave it, the OS will save it to swap when and if it needs to (= it's getting low on physical RAM).

---

This may be a misunderstanding on my part, but i thought the working was with the Leave in Memory:

- If off and the client is idle (paused due options like e.g. in use) but in between checkpoints, keep in memory/vm.
- If off and switching projects, unload the pre-empted project.
- If on and switching project, keep in memory for quicker restart when it its time again.

I cant see it unloading every-time a system is being used intermittently e.g....could as well exit the client or set the schedule to wait a long time before resuming. Maybe it should be pushed off into swap file?

sort off how I'd expect it to function without loss of progress.

One point you missed.

If off and switching projects AND the project being suspended has not checkpointed, the task stays in memory / swap file. This is so that projects that do not checkpoint frequently (or at all) will eventually complete if the machine stays on long enough.

---

BOINC WIKI

Actually i think implicitly from my first point "but in between checkpoints" it should apply to any similar condition, be it benchmarking, time slicing, high priority switching and other situations not mentioned. BOINC simply should not allow the unloading in cases where there is potential progress loss. The Leave in Memory acting to keep the jobs in situ, so with time-slice switching which mostly is at checkpoints, the resume is quicker. Personally i prefer long period and mostly have it set to either 240 or 360 minutes on multi project clients. Over time all get their portions per the resource rules (which can be long periods for those conditions to be really met).

BOINC does allow progress loss as long is it is not all the way back to (the beginning? the checkpoint where this run started?). Normally it waits for a checkpoint to switch (if the checkpoint has been within the last 10 seconds, it counts as current).

---

BOINC WIKI