I'm attached to 2 projects with resource shares of 50/50 and a connect-to-interval (cache) of 1 day. Theoretically shouldn't it take 1 to 1.x days from the time a task is sent until it reaches the top of the queue, is processed, returned and reported?

When I look at my results page it appears that tasks with 8 hr run times are reported around 1 to 1.3 days after being sent, yet my avg turnaround time shows as 2.2 days. At Einstein my ATT is computed as 1.2 days. Does anyone else experience this?

Maybe the answer lies in those annoying ghosts that I periodically get.

Maybe the answer lies in those annoying ghosts that I periodically get.

If so, that would mean that the ATT algorithm uses the deadline as the return time when a result times out. I can see the logic if it does, but my preference would be for ATT to mean the average of time the host has held work that is actually returned.

Another possibility is that this figure might be calculated like RAC, on an exponential decay basis, in which case you might still be suffering from a run of long turnrounds some time ago. That is only a guess - I don't have any evidence let alone any actual knowledge about how it works it out.

R~~