Large difference between CPU time and Run time

[BobCat13]

Host 11411 is showing quite a difference from CPU time to Run time

http://www.rnaworld.de/rnaworld/results ... stid=11411

What could be causing the difference?

Looking at the stderr_txt in the task result it appears that the task checkpoints. Could it be restoring from the checkpoint keeping the Run time but resetting the CPU time to 0 and then finishing the workunit shortly after the restored checkpoint? It is really killing the credit for any wingman that matches up with this machine.

[Txt.file]

this can happen for a few different resons. one is that boinc (or any other program) limits the cpu time. so that the app doesn't get 100% cpu time.

[mxplm]

Do you have any other CPU intense tasks running on your box? CPU time will always given to whatever process needs it before BOINC gets something.

[Ananas]

If you compare it to the wingmen, CPU time must be wrong, so it is most likely not an issue of concurrent high-load programs.

It might be a "freeze" issue as the boxes with a correct CPU time have no Freeze entries in stdout.

As CPU time is the base for the credits, the wingmen might not like this

[mxplm]

Can freeze save and restore the complete state of a process, including the cpu time? Otherwise, this problem could potentially occur on any 32Bit Linux when rebooting.

[Ananas]

I guess the CPU time is stored in the kernel task-wise. If a task ends, the kernel looses the information, if a new task is fired up, the counting restarts at 0. I doubt that Freeze has write access to this information.

There is one program though, that would theoretically be able to accumulate the CPU times of different partial runs : the wrapper.

[BobCat13]

Do you have any other CPU intense tasks running on your box? CPU time will always given to whatever process needs it before BOINC gets something.

It is not my machine. I just happened to be a wingman on a task. When I saw the low credit awarded, I clicked on host 11411 and noticed the large difference on most tasks that machine has run.

[BobCat13]

A workunit like this one is why I inquired about host 11411 and why I aborted my copy (host 1734) prior to starting.

http://www.rnaworld.de/rnaworld/workuni ... id=5294419

There is a chance somebody is going to finish that task with a claimed credit in excess of 1,000 and receive 2.50.

[Michael H.W. Weber]

There is a chance somebody is going to finish that task with a claimed credit in excess of 1,000 and receive 2.50.

This indeed appears a little strange.

Michael.