percentage completed changes when workunit resumes.

Message boards : Number crunching : percentage completed changes when workunit resumes.

To post messages, you must log in.

AuthorMessage
racquemis

Send message
Joined: 3 Dec 05
Posts: 2
Credit: 70,768
RAC: 0
Message 54464 - Posted: 12 Jul 2008, 23:09:00 UTC
Last modified: 12 Jul 2008, 23:09:21 UTC

I've noticed something strange which i was suspected for quite a while
whenever i switch off my computer during crunching, i close boinc the normal way and then switch off the computer. then when i startup bboinc again the next day i often see that there are workunits ready for upload/reporting. i didn't recal finishing any workunits the last day and normally i always report workunits before i close boinc.

Today i tryed it manually, i paused a workunit at 76% and resumed it after a while, it would immidiately after initialization report the workunit as 100% and start uploading it.

I'm wondering if anyone else is noticing this, could this be an error in the workunit, or could it be just something as innocent as a percentage miscalculation?

thanks in advance
ID: 54464 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Speedy
Avatar

Send message
Joined: 25 Sep 05
Posts: 163
Credit: 808,098
RAC: 0
Message 54485 - Posted: 14 Jul 2008, 5:18:44 UTC - in response to Message 54464.  

I've noticed something strange which i was suspected for quite a while
whenever i switch off my computer during crunching, i close boinc the normal way and then switch off the computer. then when i startup bboinc again the next day i often see that there are workunits ready for upload/reporting. i didn't recal finishing any workunits the last day and normally i always report workunits before i close boinc.

Today i tryed it manually, i paused a workunit at 76% and resumed it after a while, it would immidiately after initialization report the workunit as 100% and start uploading it.

I'm wondering if anyone else is noticing this, could this be an error in the workunit, or could it be just something as innocent as a percentage miscalculation?

thanks in advance

i think you will find that it depends how far through the model it is when you pause it. It will go back to it's most recent checkpoint & start from there. As for the work unit finishing @ 70% I think you will find that the workunit can't complete another full model be for the workunit time limit e.g 3 hours. Please correct me if I'm wrong. i hope you can understand what i am saying.
Cheers
Speedy
Have a crunching good day!!
ID: 54485 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Greg_BE
Avatar

Send message
Joined: 30 May 06
Posts: 5691
Credit: 5,859,226
RAC: 0
Message 54490 - Posted: 14 Jul 2008, 11:20:18 UTC - in response to Message 54485.  

I've noticed something strange which i was suspected for quite a while
whenever i switch off my computer during crunching, i close boinc the normal way and then switch off the computer. then when i startup bboinc again the next day i often see that there are workunits ready for upload/reporting. i didn't recal finishing any workunits the last day and normally i always report workunits before i close boinc.

Today i tryed it manually, i paused a workunit at 76% and resumed it after a while, it would immidiately after initialization report the workunit as 100% and start uploading it.

I'm wondering if anyone else is noticing this, could this be an error in the workunit, or could it be just something as innocent as a percentage miscalculation?

thanks in advance

i think you will find that it depends how far through the model it is when you pause it. It will go back to it's most recent checkpoint & start from there. As for the work unit finishing @ 70% I think you will find that the workunit can't complete another full model be for the workunit time limit e.g 3 hours. Please correct me if I'm wrong. i hope you can understand what i am saying.
Cheers
Speedy


I know modsense has a thread somewhere that discusses this.
speedy, you are correct in your statement, here is what i remember further to this discussion elsewhere.
racquemis - boinc will try and complete as many models as possible within your run time preferences. If it can not complete any more within the runtime preferences it will stop the process and make it done. As for stoping at 76% and then reporting done, I have had similar issues, it probably was reporting 76% as the last known point and trying to finish one last model before your deadline. however if it got interupted boinc may feel that there is not enough time left in your runtime when you restarted the task to complete the last model and decides to call the task done.
ID: 54490 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 54494 - Posted: 14 Jul 2008, 16:24:01 UTC

This task shows that it was ended due to too many restarts. A "restart" occurs any time the task begins from the last checkpoint. So, this means any time BOINC is restarted, and also, if you do not keep tasks in memory while suspended, any time the task is preempted to crunch something else.

Suggest you review your computing preferences and keep tasks in memory while suspended. This keeps them out in virtual memory until they are ready to run again. Thus it tends to preserve more of the work your machine has completed.
Rosetta Moderator: Mod.Sense
ID: 54494 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Greg_BE
Avatar

Send message
Joined: 30 May 06
Posts: 5691
Credit: 5,859,226
RAC: 0
Message 54499 - Posted: 14 Jul 2008, 17:51:14 UTC - in response to Message 54494.  

mod.sense - question for you how does boinc keep a task in memory if your system is powered down or restarted? I thought memory got turned off as well when you do a restart?

This task shows that it was ended due to too many restarts. A "restart" occurs any time the task begins from the last checkpoint. So, this means any time BOINC is restarted, and also, if you do not keep tasks in memory while suspended, any time the task is preempted to crunch something else.

Suggest you review your computing preferences and keep tasks in memory while suspended. This keeps them out in virtual memory until they are ready to run again. Thus it tends to preserve more of the work your machine has completed.

ID: 54499 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
racquemis

Send message
Joined: 3 Dec 05
Posts: 2
Credit: 70,768
RAC: 0
Message 54503 - Posted: 14 Jul 2008, 20:35:31 UTC - in response to Message 54494.  

This task shows that it was ended due to too many restarts. A "restart" occurs any time the task begins from the last checkpoint. So, this means any time BOINC is restarted, and also, if you do not keep tasks in memory while suspended, any time the task is preempted to crunch something else.

Suggest you review your computing preferences and keep tasks in memory while suspended. This keeps them out in virtual memory until they are ready to run again. Thus it tends to preserve more of the work your machine has completed.


that clear things up, thank you. i didn't had keep task in memory while suspended activated. but that solves the problem when the pc is powered on
It sometimes happens my pc randomly restarts or when i need to turn it off, since i can't let it stay on 24/7 unattended due to a pretty high overclock
(50% overclock on Q6600, stable)

Is the only thing i can do is just wait until the tasks finishes before shutting down.

BTW, is is bad if a workunit is reported when it's not complete, i mean for bakerlab?
ID: 54503 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1832
Credit: 119,677,569
RAC: 10,479
Message 54504 - Posted: 14 Jul 2008, 21:34:22 UTC - in response to Message 54499.  

mod.sense - question for you how does boinc keep a task in memory if your system is powered down or restarted? I thought memory got turned off as well when you do a restart?

If the system is put on standby or hibernated then the task is kept 'in memory'. If the system is shut-down then the task is restarted. Mod.Sense means that if the 'stop crunching when computer is in use' option is selected, and 'keep in memory when pre-empted' isn't selected, then Rosetta will effectively be stopped when the computer becomes active, and will restart when the computer is idle again...
ID: 54504 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 54505 - Posted: 14 Jul 2008, 22:14:41 UTC

Right, when the machine is powered off, or BOINC is shutdown, then the task only has it's last checkpoint to start at. I was attempting to show two situations where one might end up with enough restarts that the task is ended like that.

racquemis, perhaps you can undervolt or reduce clock speed so you can crunch 24/7. I mean what is the point of overclocking, if you have to shutdown periodically and lose what you were working on?

Perhaps you have other reasons for the overclocking. You might consider setting BOINC CPU preference to run at less then 100% of CPU. That can help you run cooler.
Rosetta Moderator: Mod.Sense
ID: 54505 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1832
Credit: 119,677,569
RAC: 10,479
Message 54506 - Posted: 14 Jul 2008, 22:29:47 UTC - in response to Message 54505.  

Right, when the machine is powered off, or BOINC is shutdown, then the task only has it's last checkpoint to start at. I was attempting to show two situations where one might end up with enough restarts that the task is ended like that.

racquemis, perhaps you can undervolt or reduce clock speed so you can crunch 24/7. I mean what is the point of overclocking, if you have to shutdown periodically and lose what you were working on?

Perhaps you have other reasons for the overclocking. You might consider setting BOINC CPU preference to run at less then 100% of CPU. That can help you run cooler.

or hibernate it rather than shutting down - I'm fairly sure the Rosetta threads aren't restarted on resume from hibernate(?)
ID: 54506 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : percentage completed changes when workunit resumes.



©2024 University of Washington
https://www.bakerlab.org