Message boards :
Number crunching :
Unit prioritization
Message board moderation
Author | Message |
---|---|
Shawn Steele Send message Joined: 20 May 99 Posts: 5 Credit: 7,030,927 RAC: 0 |
I'm really confused how boinc picks which seti tasks to run first. Currently I have several tasks that pretty much fit in about 3 buckets: * Several that are estimated at 1:17 and due by 7/21 * A couple that are estimated at around 2:54 due on 8/15 * Several estimated at 3:19-3:28, due by 8/24-8/25 There's pretty much zero chance that any of them "won't get done" if I just run the computer, but Boinc picks several that aren't due for 1 1/2 months. Sure, they're a little longer, and it should be able to do those, then scoot back to pick up the ones with a month earlier deadline, but why? What Boinc doesn't know is that I might want to go on a summer vacation. In which case the computer could be off completely for 2-3 weeks. In which case there's still no real problem doing all of the work before the deadlines, *IF* it starts with the soonest work first. If it insists on doing August work before July work though, then the July work could be at risk. I've seen threads that say "Boinc's smart, let it do it's thing", but AFAICT the only way BOINC can be messed up is if I turn off the machine for a vacation. And, if I do that, then it's current algorithm has a great chance of missing deadlines unnecessarily. Can someone explain what I'm missing? Maybe other projects have more trouble scheduling units? |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
Unless something has changed, BOINC always goes first in first out(FIFO) for work completion. In the unlikely event that your computer is likely to miss a deadline it will start working the shortest deadline first. In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
Actually I would've thought that you would've asked about why your rigs keep trashing AstroPulse v6 v6.08 (cal_ati) work than worry about report deadlines. Cheers. |
Mike Davis Send message Joined: 17 May 99 Posts: 240 Credit: 5,402,361 RAC: 0 |
Actually I would've thought that you would've asked about why your rigs keep trashing AstroPulse v6 v6.08 (cal_ati) work than worry about report deadlines. +1 |
Les Send message Joined: 20 May 99 Posts: 53 Credit: 21,062,237 RAC: 18 |
Unless something has changed, BOINC always goes first in first out(FIFO) for work completion. In the unlikely event that your computer is likely to miss a deadline it will start working the shortest deadline first. I have to disagree - whenever the estimated duration goes haywire such as AP estimated to take 30-40 times longer than they ever have, the tasks running as high priority seem to be more randomly selected. On my system, a few weeks ago tasks with deadlines two weeks away were skipped and those with deadlines four to five weeks away were running in false panic mode. I do not have screen captures to show this but I am sure that others have often reported similar issues. I have also noted that often tasks with much earlier deadlines were often suspended even if mere minutes of processing remained in favor of poorly selected "high priority" tasks with deadlines weeks later. |
Gatekeeper Send message Joined: 14 Jul 04 Posts: 887 Credit: 176,479,616 RAC: 0 |
Unless something has changed, BOINC always goes first in first out(FIFO) for work completion. In the unlikely event that your computer is likely to miss a deadline it will start working the shortest deadline first. IIRC, the logic that BOINC uses in "high priority" is really simple and more or less obvious: If BOINC thinks that a certain workunit or group of units won't finish by deadline, it'll run them in HP until the time calculations drop below "panic" level. It makes sense that such units would be the ones with the farthest out due dates, BOINC thinks they won't get done in time. |
Shawn Steele Send message Joined: 20 May 99 Posts: 5 Credit: 7,030,927 RAC: 0 |
"Unless something has changed, BOINC always goes first in first out(FIFO) for work completion" But BOINC has no clue when I'm going to take a vacation :) So it can't factor that into its calculations. So I would much rather do "soonest deadline" first rather than FIFO. Most of the deadlines seem to be "pretty long", so that'd take care of most vacation scenarios. Unless the backend needs them back in FIFO order for some reason (but if it wants them sooner, wouldn't it provide sooner deadlines?) |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
or do what many of us do and just leave it running while you are away. Alternately, you can set it to No New Work and let it run down until you leave. In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
Vicki Send message Joined: 30 Nov 01 Posts: 65 Credit: 1,640,576 RAC: 46 |
"Unless something has changed, BOINC always goes first in first out(FIFO) for work completion" I agree, setting no new tasks a week or 3 before your planned holiday is 1 way around it. The other thing you can do <which I forced Bonic to do rescently, when I had 18 work units all due within 2 weeks & 30+ other workunits due in october> is suspend all tasks except those with the shortest deadlne until they are completed, then resume the other tasks in stages according to dealine, all of which can be done from the tasks page from within Bonic. To make it easier, you can have Bonic list all tasks by due date by clicking on the column header. Hope that Helps. Rae A city destroyed by an earthquake is an opportunity to Rebuild, redeign & make it a better place to be. Better, stronger, faster like the 6 Million Dollar Man |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
You can force this 'Unit prioritization' by: Switch between tasks every XYZ minutes (play with values 5000-15000) Depending on your habit set it locally or here: http://setiathome.berkeley.edu/prefs.php?subset=global  - ALF - "Find out what you don't do well ..... then don't do it!" :)  |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.