I'm not sure what algorithm BOINC uses to allocate work among different projects. I would have thought that if I gave a score of 100 to each project, it would allocate an equal number of CPUs to each one. Not so! Universe@home consistently uses ALL the CPUs, leaving SETI@home and climateprediction.net waiting forever for a turn.

It appears that the deadlines figure into the allocation of resources, and Universe@home uses MUCH shorter deadlines than these other two projects. A newly downloaded task might have a deadline as soon as a week (which may not even be achievable by some of my older computers). Other projects use longer, more realistic deadlines, and therefore get NO CPU TIME AT ALL. I have to individually suspend Universe@home tasks in order to give the others a chance.

Three questions: Am I correct about the way the allocation of CPUs to tasks works (deadline dependent, among other things)?

Can you make the deadlines 2-3 times longer, so my older computers have a shot at finishing a task on time?

And most importantly, can you adjust things so that Universe@home doesn't hog all the resources? When I tire of playing scheduler, the easiest solution will be to remove this project from my BOINC list.

---

I haven't had a problem with Universe and: LHC/ATLAS, Einstein, Rosetta, Milky Way, or any other project. Maybe you have temporarily downloaded too many work units, and BOINC has to give Universe priority to finish on time. That can happen if the time estimates of some projects are misleading, and BOINC downloads too many. There are probably other scenarios where it can happen. I think it will straighten itself out as BOINC corrects it time estimates.

Another possibility is that you have set too large a work buffer. Reduce it to the defaults (0.1 + 0.5 days) to see if that fixes it.

The client runs the tasks and has control over which task runs next. Any project has no control over that. The resource share are in terms of many weeks. Until you run all projects set to DL work for a long time the client will try to balance it out. So a new project to you will get all the resources to 'catch up' to the processing time of the other projects on your PCs.

If these take too long on a slow PC then you'll have to run something else.

There is actually a way to correct for that, which I do on all new installs (both Windows and Linux).
You create a "cc_config.xml" file. Use a text editor, such as notepad, and then use the "save as" function to save it as an ".xml" file (not a .txt file).

Here it is:

<cc_config>
<options>
<rec_half_life_days>1.000000</rec_half_life_days>
</options>
</cc_config>

Then, you place it in the BOINC Data folder, and restart BOINC (or reboot). It shortens the time to correct the time estimates to about a day or two.

Thanks so much for that. I tried it, but the Universe@Home tasks keep preempting the SETI ones. I have to suspend them to keep them from using all the CPUs.

I used SETI@home for years before adding Universe@Home as a project; are you guys saying that BOINC is trying to balance the workload over all of my computer's history, not just in current time? Yikes!

I used SETI@home for years before adding Universe@Home as a project; are you guys saying that BOINC is trying to balance the workload over all of my computer's history, not just in current time? Yikes!

Over about 20 days to a month I think. It is too long, but not infinite.