Resource share for WU downloads

Hi all. I was just wondering if anyone knew how BOINC or maybe it is BAM decides which project has the highest priority in regards to WU downloads when all are set to the same percentage of resource share?. I frequently get this message in my BAM event log '' Not requesting tasks: don't need (CPU: not highest priority project; AMD/ATI GPU: not highest priority project)'' I have tried increasing percentage of resource share to see if this makes a difference but it seems not to. Apart from zeroing out other projects or setting to no new tasks, is there a way of ensuring my favourite projects always get work? Is there a setting I have not yet discovered perhaps?
Thanks in advance. :)

I have the same problem, and no real good answer unfortunately. I think if you increase the resource share for your favorite projects (or decrease for the least favorite ones) it might help, but it's not a full solution.
Once Boinc fills up the queue to how many days work you set it to, I think it won't get more work even if a higher priority (resource share) project is active, until it at least is able to clear some of the queue. The quick solution of course is aborting WUs of least favorite projects so that the queue is cleared and boinc fills it back up with the higher priority projects.
They have an article on the scheduling policies on their wiki here: https://boinc.berkeley.edu/trac/wiki/ClientSched
I don't fully understand it, but I think the gist of it is that BOINC commits to the work it has in queue and if it was filled with WUs from least favorite projects first (for example because best project XYZ was out of work at the time) it will have to go through those WUs first before getting more from the favorite projects.

Probably the best way to go about it would be to have the queues to 0 or smallest possible value so that BOINC does a quick turnaround and constantly checks the high priority projects for work. The downside to that is that if projects go out of work or you lose internet connection your machine might go idle and CPU cycles will be wasted until the issue is fixed.

Some projects can overburden you when set to 100%, others need 10,000% to get any work at all.

Hmm, yes thanks Neo, I may clear it all out and set to 1 days work only and see what happens. I often have wondered though that if I do that and I get a big WU that lasts 4 or 5 days or longer, will that then stop any other project from getting Wu's?
Thanks Dirk also, I have also found some projects go overboard with the no. of Wu's downloaded and have had to abort as they wont finish before their deadline.
Do you all use an account manager or do you control projects etc yourself? I am using BAM from boincstats currently. Perhaps those of you using linux/Ubuntu have more control than us windows users!

I've never noticed difference in control using Linux vs Windows or using BAM! vs managing myself.
BAM! is great when setting up a new system, but there is stops, for me.

I use only BOINC's own Client in Windows on the few machines I have, if I had a large fleet I might consider a better option.
I believe that if you happen to download a very long WU (like Climate prediction for example) you will be stuck with it for how long it takes even if the buffers are set to 0 days. I think BOINC commits to the work it receives to the fullest extent it can (unless you abort it)

Quote:

---

Originally Posted by NeoGen

a very long WU (like Climate prediction for example)

---

I have a RNA WU here that is running for three days and wants another 94 days more...now. I hope to finish it in 2018, somewhere.

Quote:

---

Originally Posted by Dirk Broer

I hope to finish it in 2018, somewhere.

---

:sad5::sad5::sad5:

It now has run for 14 days, wants another 90 days more AND is supposed to be finished next thursday.
Fortunately they keep pushing back the deadline, but my holidays may prove fatal for this WU.

Can you suspend it?
if so, go to the setting
Computing Prepferences >> Disk and memory >> "Leave non-GPU tasks in memory while suspended" and uncheck it so it suspends to disk.
Then don't forget to suspend the project via the projects tab.
Then when you come back from holiday check your tasks to see if it has timed out or if it is still pending.
Then you will know whether to abort it or resume it.

Alternatively - if it is not a Wu that you can normally suspend without it re-starting - try putting the computer into hibernate.
Sometimes that works when the other does not.

Too late, It's switched off now and won't be re-started until the 2nd of August....and by the time I've come back Bok will hopefully have restored the working of ex-Goofyx/now-GCC subprojects and badges, plus the badges for Climate2 and Amicable numbers. If anyone can get the CERN boys so far as to adopt the GCC solution for the export stats for both LHC@Home and LHC-dev@Home, I'll be one happy camper upon return.

Have a great break. Hope the weather is kind to you :)