LHC@home

Deadlines are assigned by the individual projects, not by BOINC. ClimatePrediction, for example, has a deadline of next year. ProteinPredictor has a deadline of this week. Under the new 4.43 "deadline" scheme, PP will always get priority - and it did. After I upgraded BOINC to 4.43, PP was taking over all of my machines.

In the online preferences, I have assigned equal weight to all projects. I expect BOINC to honor MY decision, and automatically switch between projects every 2 hours, as I have instructed.

This new "deadline mode" is too smart for it's own good, IMO.

I have reverted back to 4.25 until BOINC drops "deadline" mode -or- better yet, gives me the option of using it or not.

--- bt

The theory is that yes, P@H WUs will get processed first due to their deadline, but while they are running, BOINC is keeping track of how long they run. When they are complete, BOINC will not download any more P@H work until the other projects are able to catch up on CPU time, according to your resource share.

So your resource shares are being met, just over a longer term.

---

> The theory is that yes, P@H WUs will get processed first due to their
> deadline, but while they are running, BOINC is keeping track of how long they
> run. When they are complete, BOINC will not download any more P@H work until
> the other projects are able to catch up on CPU time, according to your
> resource share.
>
> So your resource shares are being met, just over a longer term.
>
>

This is exactly what is happening with the new shedular and it kicks ass in my opinion... MUCH better than the timeslice system previously implemented... I have two boxes running four projects each, all projects have 100% share 100/4 = 25... 25% per project, per box. I haven't missed a deadline yet!! Short of the ones that the client lost track of during upgrades/restores.

---

<img src="http://boinc.mundayweb.com/one/stats.php?userID=444&amp;trans=off">