Message boards : Number crunching : New WUs being started before older ones are done
Author | Message |
---|---|
Rick Spies Send message Joined: 5 Aug 10 Posts: 5 Credit: 1,718,814 RAC: 0 |
I pretty new to Rosetta (but have been on the SETI project for >10 years). I am using a dual core PC for Rosetta. I've noticed that Rosetta will start a new pair of WUs before finishing other WUs. For example, right now I have six WUs in various stages of completion: 27% 21% 11% 10% 9% 5% Rosetta always works on them in pairs, as shown above, switching from one to another or starting a new pair at any time. QUESTION: Is it normal for Rosetta to begin new work units before completing others? This does not happen on SETI. Looking at my WUs, I can see this work shifting is jeopardizing completion of partially-processed WUs before their reporting deadlines, wasting a lot of time and CPU cycles. WUs are not being worked on in report deadline order either. Any insight would be appreciated. Rick Portland, OR ================================================= There is never a wrong time to do the right thing |
Greg_BE Send message Joined: 30 May 06 Posts: 5691 Credit: 5,859,226 RAC: 0 |
Don't know why it does that Rick. I have the same thing going on on my system Of course it is in groups of 4 with my quad core cpu. I have 3 sets of 4 in progress! Weird. Now 1 of the tasks has to run in "high priority mode" |
Mod.Sense Volunteer moderator Send message Joined: 22 Aug 06 Posts: 4018 Credit: 0 RAC: 0 |
Rick, I hear people report such things every once and a while, but it's not all that common. I would hope they would run in expiration order as well. What BOINC Manager version are you running? It is really the BOINC Manager that decides which tasks to schedule when, not the Rosetta application. Perhaps there is a fix in a newer version. Rosetta Moderator: Mod.Sense |
Rick Spies Send message Joined: 5 Aug 10 Posts: 5 Credit: 1,718,814 RAC: 0 |
Rick, I hear people report such things every once and a while, but it's not all that common. I would hope they would run in expiration order as well. What BOINC Manager version are you running? It is really the BOINC Manager that decides which tasks to schedule when, not the Rosetta application. Perhaps there is a fix in a newer version. I'm running BOINC v6.10.58, which is the most recent version. To be a bit more complete, I'm running Win7 x64 on an Intel Core2Duo proc. This has been a problem since I started Rosetta in August. It's not a problem with SETI@home. If there is anything I can do to help you troubleshoot this problem, just ask. ================================================= There is never a wrong time to do the right thing |
mikey Send message Joined: 5 Jan 06 Posts: 1895 Credit: 9,208,737 RAC: 3,249 |
Like Mod.Sense says, this is not a project 'thing'. The manager, not the science application, decides when the tasks are run. It could be tasks on SETI do not run as long. Maybe the SETI-tasks are completed before BOINC decides its time to switch to other tasks. But since it is a Boinc thing it must get the info from someplace, so I wonder if Rosie did something different to its units. The OP says his Seti units are not doing that, so it COULD be a project thing, just a unit creation project thing, not a user fixable problem. |
Mod.Sense Volunteer moderator Send message Joined: 22 Aug 06 Posts: 4018 Credit: 0 RAC: 0 |
I can only suggest reporting this as a quirk of the .58 release then. And offer to you that I concur, that it's doing no harm so long as the tasks are completed and reported back before their deadlines. Rosetta Moderator: Mod.Sense |
Murasaki Send message Joined: 20 Apr 06 Posts: 303 Credit: 511,418 RAC: 0 |
A little search on the net shows this bug has cropped up before in earlier 6.10.X versions for both the Collatz and Einstein projects. It is definitely not a Rosetta-specific glitch. |
mikey Send message Joined: 5 Jan 06 Posts: 1895 Credit: 9,208,737 RAC: 3,249 |
A little search on the net shows this bug has cropped up before in earlier 6.10.X versions for both the Collatz and Einstein projects. It is definitely not a Rosetta-specific glitch. So it is a Boinc thing! |
Bill Hepburn Send message Joined: 18 Sep 05 Posts: 14 Credit: 14,953,680 RAC: 4,081 |
One of the things you might do is set "Switch between tasks" in your Computing Preferences to some large value (I use 720 minutes) instead of the recommended (default) 60 minutes. That way, most tasks have time to complete before BOINC switches applications. It switches when tasks complete anyway, and I almost never see a partially completed task sitting around. I haven't seen a down side do doing this, but there may be one. |
Greg_BE Send message Joined: 30 May 06 Posts: 5691 Credit: 5,859,226 RAC: 0 |
@Bill thanks for that tip on switching time on tasks. I put mine up to 240 mins from 60 mins and it looks like Boinc will process the tasks in the order they are due to report instead of all the mickey mouse stuff it was doing.
|
mikey Send message Joined: 5 Jan 06 Posts: 1895 Credit: 9,208,737 RAC: 3,249 |
I am slapping my forehead because I did not think of that! Darn I hate 'senior moments'!!! VERY nice one!!!! |
Message boards :
Number crunching :
New WUs being started before older ones are done
©2024 University of Washington
https://www.bakerlab.org