I'm running a single core machine under Windows 7, originally under BOINC 7.2.4 and now (after a complete uninstall and deletion of program and data folders) 7.3.11. I saw under 7.2.4, and am still seeing under 7.3.11, a bit of a problem with scheduling.

I am currently running two projects, PrimeGrid and CPDN, both with 50% share. After the installation of 7.3.11 I attached to PrimeGrid and a task was downloaded. I then attached to CPDN and a task for that project was also downloaded. The PG task ran to completion and the CPDN task then started - and that is the current state after 15 hours - no hourly task switch takes place and no more PG tasks are ever downloaded.

If I suspend the running CPDN task then another two PG tasks are downloaded, indicating that I am attached to the project correctly. FWIW, the CPDN task should take about 200 hours and the deadline is March 2015.

The thing that looks wrong to me is the project scheduling priority values. If I monitor (i.e. keep opening and closing the project properties dialog) the PG priority occasionally alternates between -0.24 and -0.48. However the CPDN value alternates between about -1.0 and -77.0 (there are small 0.01 variations). NB This is while there is just the CPDN task in progress and no PG tasks are being requested. I would guess that is why no more PG tasks are requested?

Can anyone tell me if these scheduling priority number are what I should expect and, if so, what else I should be looking at to correct the issue.

Ian

---

The huge difference in computing times between CPDN and pretty much any other project is going to give priority to CPDN, because BOINC is worried about getting it done in time. However, it's also supposed to make sure that your total computing time is scaled to resemble something like your resource share. So, if you complete a CPDN unit, you should then get a bunch of work from your other project(s) to balance it out.

---

My Detailed BOINC Stats

Basically with that CPDN Wu, your cache is full, Boinc won't ask for work again until it needs it, or is far enough though it that it feels that another project deserves some time.

Claggy

It's the "another project deserves some time" bit that is confusing me. I have the two projects on 50% share so I would have thought BOINC would allocate some time to PG even though it has a long CPDN task in progress.

It's not as if the CPDN task is in any danger of overrunning ...

Currently done 18% in 36 hours so needs 164 more to complete.

BOINC "guesstimate" is 75 hours to completion.

Time before task deadline is 8280 hours.

One thing I forgot earlier, if I tell PG to update the log shows ...

"Not requesting tasks: don't need (buffer full)"

... even though there are no tasks in the PG buffer.

I'm not going to stress about it, I can easily micro-manage, but the large changes in the scheduling priority does make me think something is not _quite_ right.

---

One thing I forgot earlier, if I tell PG to update the log shows ...

"Not requesting tasks: don't need (buffer full)"

... even though there are no tasks in the PG buffer.

There is no such thing as a PG Buffer, The Computing preferences is a Global preference, ie, it applies to all projects together, not each project separately,
Say you have your cache settings set to 2 + 0 days, If the total guestimate for how long all your remaining work is greater than two days then your cache is full,
Boinc won't ask for work from any project because it doesn't need it, once it does need it it'll then ask for work from the project(s) with the highest priority.

Claggy