With projects that are in upstart and have highly volatile task durations like this one [follow the forum], it's best not to set the cache/buffer too high... just a day or 2 max. Seeing on my machine the range of 23 minutes and 1:31 hours, a variance of over 4x, the potential on wrong estimates causing buffer overfill is quite high. There's an FAQ in the Start Here forum explaining the "Totally Wrong" Times to Complete projection.

ciao

PS, new projects could throw fits though beta tested, and this one was extensively. Just keep the buffer smaller... there's for most projects more than enough work to go around.

And,

I think I need to add a project, for only HCMD2 shorties has let my quad run up against the daily quota wall.

2009/05/14 13:33:24 World Community Grid Message from server: No work sent
2009/05/14 13:33:24 World Community Grid Message from server: No work is available for Help Cure Muscular Dystrophy - Phase 2
2009/05/14 13:33:24 World Community Grid Message from server: (reached daily quota of 320 results)


Reported to the techs as there are probably more who only do this project and are potentially running up against the daily download limit.

Added: The shorties I got in the hundreds were all of the PP1BA-RADIA type.

edit: Typo

Had to abort a couple of downloads as they kept bouncing and telling on wrong sizes, but no computation errors so far. Got enough in cache I think to last through 0:00 UTC, but they run much quicker still than the 9 minutes predicted... oops, now 7:53 minutes after last completions. BOINCview claims present stock is worth 1:06 days... on 4 core that's barely 6 hours.

That's not selfish, that's comprehensible...!!!

As long as these WUs are as short as they pretend to be or even shorter it's a problem only for your daily quota and this is something that knreed can arrange quickly if he wants to.

But cross your fingers to avoid the Pseudo very short WUs!
I have received about 200 of them in no time this morning while I was looking elsewhere and this device has now far more work than I wanted for it.

I can understand the difficulty to chop WUs the size you want, but with the awkward scheduling algorithm of Boinc I would prefer that the estimated fpops be overestimated rather than this ridiculously low one.

In this morning's batch the really very short ones were very rare and still three times longer than estimated, and most of those which have been crunched since then are actually of the normal short size (about one hour vs the estimated 8 minutes).

Good luck chaps! Jean.