Message boards :
Number crunching :
"High-Priority" tag?
Message board moderation
Author | Message |
---|---|
Send message Joined: 6 Sep 09 Posts: 3 Credit: 81,863 RAC: 0 |
It's been quite a while since I've done any 'crunching' for this project, and I do have a question ... The WU (an N-Body Simulation) has a high-priority tag on it and an estimated time in excess of 156 hours and a ten day window to complete. Early response of the BOINC manager suggests the estimated time to be high by a factor of 10. None of my other projects are in time crunch, so Im just going to let it run. Is this an anomaly limited to my machine alone, or has thee been issues of this in the past? |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
See any of the other numerouts posts on this. Why no mention is made on the site, I don't know. N-body tasks have a poor estimate time. The best way is to let it run a few and calculate it out to see if you want to run it. They are all like this. I have had 2hours to 80+ hours on these tasks. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 6 Sep 09 Posts: 3 Credit: 81,863 RAC: 0 |
Thanks for the quick reply there, Bandit ... As I mentioned previously, none of my other works are in time pressure, so I'm going to let it go (should be finished in the morning). Then, I'll re-evaluate |
Send message Joined: 19 Jul 10 Posts: 595 Credit: 18,964,932 RAC: 5,531 |
Once you have 10 valid WUs for the n-body application, the "server side DCF" kicks in, and the estimates will be (or should be) close enough to the real runtimes, so they should not run in high-priority mode all the time (after the client side DCF has adjusted to the new estimates). |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
Once you have 10 valid WUs for the n-body application, the "server side DCF" kicks in, and the estimates will be (or should be) close enough to the real runtimes, so they should not run in high-priority mode all the time (after the client side DCF has adjusted to the new estimates). I've run dozens and it only dropped from ~850 hours to ~450 hours even though my longest has been ~80 hours. Mw has the n-body really overestimated. The ps-tasks corrected to the right time long ago. @Ocean: MW has almost always ran in high priority mode for me. It doesn't bother me as I seem to have my computer set how it works for me. If you don't want to run the n-body you can set yourself to only run ps tasks, which for now are shorter tasks. I believe you need an app_info file for that. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 19 Jul 10 Posts: 595 Credit: 18,964,932 RAC: 5,531 |
I've run dozens and it only dropped from ~850 hours to ~450 hours even though my longest has been ~80 hours. Mw has the n-body really overestimated. The ps-tasks corrected to the right time long ago. ??? On the applications page for your only host I see just one completed task for "MilkyWay@Home N-Body Simulation 0.66". OTOH, with 2 consecutive valid tasks and only 1 completed, it looks somehow broken and might explain why the server side DCF didn't kick in yet. |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
That is most likely because I had to restore my computer last week. It only give a few tasks completed total, and that is about all I have done this week since MW came back online. I haven't had any invalids since the new apps came out. I do check each task to see they validate and I get credit. This has to do with the settings of MW/Boinc since numerous other people keep posting this very question about multi-hundred hour tasks. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 19 Jul 10 Posts: 595 Credit: 18,964,932 RAC: 5,531 |
This has to do with the settings of MW/Boinc since numerous other people keep posting this very question about multi-hundred hour tasks. Wrong estimates for a new host or app are not suprising, but if that stays that way after 10 validated results, than that seems to be indeed some milkyway/n-body problem. |
©2024 Astroinformatics Group