Is it normal for tasks from numerous projects to run at high priority when upgrading to a new BOINC version? I was ofline for a while, and upgraded from 6.4.7 to 6.10.18. Shortly thereafter I then upgraded to 6.10.43. I am not sure if I was using 6.10.18 long enough for it to "stabilize". Is this a simple function of LTD (long term debt) and therefore nothing to be concerned about? I have seen that when some tasks jump to high priority, they suspend other tasks before a checkpoint is reached, and therefore I run the risk of losing work already performed. Note: report deadlines do not seem to be the issue.

---

Note: report deadlines do not seem to be the issue.

That highly depends on when those deadlines are and how many tasks you have that have that deadline (or further away).

It also depends on which projects you are attached to, how long you were off line, how long BOINC runs per day, how much work you have in total for all projects, etc. etc.

JM7 wrote an excellent reply to someone on the Seti forums in answer to why BOINC doesn't run all work by deadline, which explains this pretty much, on a one project with multiple applications (and deadlines) principle:

Assume a project with varying crunch times and varying deadlines. (similar to astropulse and SETI). Assume that the longest crunch time is longer than the shortest deadline. Assume a dual CPU system. Assume a queue that is longer than the shortest crunch time (at least 2 of these tasks on hand at all times).

Now download 1 of the tasks with the longest crunch time (and a deadline to match). The rest of the queue will be kept full of the short tasks. Now the short tasks will be run back to back until the long task does not have sufficient wall time left to complete. The long task will ALWAYS be returned late if EDF is always used. The long tasks have to get some CPU time occasionally so that they will get done by deadline.

So, within a project, the approximate order of download is the measure used (ties are run EDF, but that is OK). This would normally give long running tasks priority for the CPU over more recent downloads - just so that they can finish.

EDF in this case is Earliest Deadline First, or the High Priority. You can easily extrapolate this to multiple projects, multiple deadlines, multiple applications (per project).

---

I guss upon further reflection, report deadlines are probably the proximate cause. What cofuses me, however, is that there seem to be no instances where any report deadline is even close to being in jeapordy. ( by my admittedly rudimentary calculations anyhow ) I was offline for a good six months unfortunately, and therefore had no tasks in any state of progress.

---

Unfortunately, BOINC can't (yet) predict the future, so it can only derive assumptions from the past. It may well be assuming that, because you switched the computer off for six months, you might switch it off again before the tasks are finished. (Question: could we teach BOINC the concepts of 'summer' and 'winter'? Many volunteers have a seasonal variation in the level of contribution they are able to make).

If you look at the 'Computer Summary' page for your machine on the website of any project you're attached to, you can see BOINC's estimate of "% of time BOINC client is running" (towards the bottom of the page). If that figure is very low, then it's not surprising if BOINC tries to make urgent use of whatever little time it expects to have available.

Thanks, guys. Had to go to work, hence the long lag time here. Without taking the time to check a specific project, I have noticed in the past what I consider to be a strangely low value for the indicator you mentioned considering I run BOINC ( and my computers for that matter ) at 100% CPU 24/7. Does BOINC consider my computer not to be running when I'm doing other tasks such as reading e-mail, etc.? This could very well be part if not all of my issue.

---