Hi all,

how is the fair scheduling with boinc 7.0.x supposed to work when there are projects involved which have different queue strategies? Currently, I see that projects with block queuing (WCG, POEM, Rosetta, QMC, ...) will download a maximum of tasks to fill up the buffer, projects with no queuing (T4T) won't load new tasks and intermittent projects (LHC1) won't load new tasks either.

What would be the ideal work buffer setting (min/max value) in such a scenario that all projects get an equal chance to download (and then execute) tasks?

Cheers,
Rob

According to this new scheduling process the request will only take place when the min boundary of the work cache is hit. However, intermittent projects like LHC1 most likely won't have at this single point in time tasks available. And when new tasks appear then the requester won't ask for it because its work cache is above the min boundary. I think this approach discriminates those projects and only systems which are tied exclusively to such projects will work on their tasks which shouldn't be general goal. Maybe it is possible to allow requests at any time for projects which are below a certain threshold of the duration_correction_factor or whatever value is responsible to calculate the scheduling debt.