dcs1955,

It is random on when the work is available on the feeder and when your machine requests work. There isn't anything extra tying results to specific hosts.

Thanks,
-Uplinger

If a machine is requesting work often enough it will get both projects, sooner or later. If not, it will not.

Mike

Your success-rate will by little to none if ONLY waiting for ARP1 jobs. You really need to run another WCG science on the side to keep the scheduler checking in with the server. If not, the connection interval will increment to as much as 24 hours, IIRC.

In my case I run all sciences on the side, each with restricted threads use, the sum though making sure no threads go idle. Doing about 55-60 jobs a day on the machine means that 55-60 times a day, at random times, the server is queried for new work. Since all the other sciences are topped out (Device Profile max allowed settings), it is in effect each time asking for a substitute for the reported task, say an MCM1, PLUS ARP1 as that is always short on the client.

Went dry 2 days ago for that one, then last night landed 4 and a 5th this morning. That will keep that 1 thread busy for the next 96 hours meaning 96 times 55 to 60 times a chance to hit on a slot with the chosen one in it.

Since the scheduler cycle after a reported task & 'no-work available' starts anew at 2 minutes, 4 minutes, 8 minutes, whatever the increment rate is, my client really connects many times more than the 55-60 completed results, upping the chances further. In effect, except for the few days, the machine has been crunching ARP1 on 1 thread pretty much non-stop since project launch, that's basically 1 out of the 2016 that gets sent out every day, since the computing time is somewhere between 21-24 hours on this host. It's though the average is 1 per day, the last time landing 4, it went without a resupply for 120 hours. And this is with the servers programmed random distribution to give all a more equal chance. I run 8 threads, some run hundreds and upwards, so their chances are greater to at least catch a few more. That's OK.

Mike, that ratio has a lot of MIP running simultaneously. Since they slow each other down, I recommend running them sparingly.

Dayle

Up to 3 mip on an 8 thread machine is ok. Likewise arp. However I check occasionally to make sure they are not likely to finish at the same time by suspending the later one for a few minutes to keep them apart. Like a car driving too close to the one in front should be braked so they don't crash.

Spreading them out helps with the random availability.

Mike