I notice that PrimeGrid has some pretty tight report deadlines for smaller work units. What puzzles me is that sometimes these work units are in danger of not being finished, but BOINC is running other workunits for the same project, downloaded later and with later deadlines, that have a much higher remaining time.

Why doesn't the scheduler attempt to just finish the task with the closest deadline, at least within a given project?

Thanks

---

It will try to do that, but also do so for all other projects you've got work from. The added problem is, what if there's two projects with likewise deadlines that'll run next? When will they be able to get their work in by deadline?

And the million dollar question is, when are you going to be interfering, deciding for yourself that BOINC should have work from all projects, so you suspend all others to force PG getting a full cache. Then after you allow the rest of the projects to continue, you then find that PG is again being run solely on its own... and then you'll complain more. ;-)

It's best to leave things be. There will be a scheduling fix or X in 6.12, mostly towards the GPU.

---

Thanks for the input.

It seems PrimeGrid makes their deadlines unreasonably short. The project downloaded another 11 work units with an estimated 191 hrs each, with a deadline of less than a week away. I have only 8 cores, so it seems to me there's no possible way all these work units could be completed in time even with 100% of all CPUs for the duration.

I'll probably cut back on PrimeGrid to avoid wasting a lot of CPU time.

Thanks

---

Just in case....
Dis you set the preference of
network usage -> connect abount every ___ days
to 0.0??

Then BOINC doesn't stockpile WU on a per project basis.
If you have 8 cores, then it should get 8 per project.
(At least that is what it does for me...)

Best Wishes....
Jay

That depends on the "additional work" value. If that's also 0.0, then you do get 1 task per CPU/core. But any value higher than 0.0 and those amounts go up.

---

I had the additional work value set to the max in case of a network outage. I just changed it to 1, so maybe that will help. Thanks for the suggestions.

---