On the individual machine, if you have a 7 day deadline workunit and a 1 day deadline workunit, then yes, it would probably get processed before the 7 day deadline. However, on the server side, the scheduler (feeder) is not full, so we don't have 7 day deadlines being sent out anymore.

Thanks,
-Uplinger

BOINC is hardcoded to process anything immediately when in 24 hour deadline. Of course if there are more than there are resources, they'll be queued. In my case, allow 6, but only permit 2 to process concurrent. All still get delivered in time.

Edit: Btw, you may notice that if you have more in buffer than allowed to process, other threads also start processing in high priority to the tune of equal to the number of FAHB you have up to all threads. In the above example. with 2 processing and 4 forced to wait, a total of 6 thread will switch to HP (7.14.2)

edit: spell devil.

Basically. shorter deadline projects jump the queue ahead of longer deadline projects unless you have virtually no queue.

There are many of us keeping the queue to a minimum in order to try to get arp units.

Another variable is the crunching time of the units, arp being much longer than the others.

Then there is the supply availability and finally the ability of the scientists to process the results.

All these factors should be taken into account to optimise the speed of the projects.

Mike

I am limiting the amount so these don't chew up all of the threads. I can get everything else done in this short queue by giving FA@H2 half of the threads. I don't have any problem getting a set done in a day, and sometimes get a second set or a third set with a new deadline for each group. I have seen generations as high as 28, but they seem to be kind of random, not progressing through the day. I am sure I pick up a mixture of some that finished quickly and some a little slower.

I'm hardly getting any work units at all. It's been extremely sporadic.