WU 6328245
-
Seeker2000
- Idle-Sammler

- Beiträge: 3
- Registriert: 07.12.2012 20:04
WU 6328245
Hi!
I'm running the WU in question now for more than 4 and a half days. It's been on 100% now for around 24h.
I find it weird, that the reference system (R.S.) has finished it 5 days and few hours. I've completed a couple of hundred WUs until now and they were ALL valid and I had no problems.
The R.S. was always (without exception) a 'crude' guide for how long will it take me for a particular WU to finish. Usually this number was between a maximum of 10 times faster than the R.S., but it could be as low as only 3 times faster or in exceptional cases even a little less (e.g.: only around 2.5x faster).
But NEVER until now the 'completion factor' has fallen as low as ONLY 1.18 and is falling as the computing continues.
Did I miss something, are my assumptions of using the R.S. as an 'indicator' of how long will the the computing for a WU last wrong or is there a problem with the WU itself?
My computer ID is 20083 and this is the link to this WU stats:
http://www.rnaworld.de/rnaworld/workuni ... id=6328245
Thank You for the answer.
Denis
I'm running the WU in question now for more than 4 and a half days. It's been on 100% now for around 24h.
I find it weird, that the reference system (R.S.) has finished it 5 days and few hours. I've completed a couple of hundred WUs until now and they were ALL valid and I had no problems.
The R.S. was always (without exception) a 'crude' guide for how long will it take me for a particular WU to finish. Usually this number was between a maximum of 10 times faster than the R.S., but it could be as low as only 3 times faster or in exceptional cases even a little less (e.g.: only around 2.5x faster).
But NEVER until now the 'completion factor' has fallen as low as ONLY 1.18 and is falling as the computing continues.
Did I miss something, are my assumptions of using the R.S. as an 'indicator' of how long will the the computing for a WU last wrong or is there a problem with the WU itself?
My computer ID is 20083 and this is the link to this WU stats:
http://www.rnaworld.de/rnaworld/workuni ... id=6328245
Thank You for the answer.
Denis
-
yoyo
- Vereinsvorstand

- Beiträge: 8170
- Registriert: 17.12.2002 14:09
- Wohnort: Berlin
Re: WU 6328245
The reference system hasn't finished the WU. We don't said this. If this would be case there would be no need to distribute this WU. The runtime (based on our reference system) is just an estimation (and sometimes a bad one).
It is just a ballpark number for you, but not an exact one.
yoyo
It is just a ballpark number for you, but not an exact one.
yoyo
-
Seeker2000
- Idle-Sammler

- Beiträge: 3
- Registriert: 07.12.2012 20:04
Re: WU 6328245
Ok.
In this case my assumption was wrong. But nevertheless the 'method' I described in my previous post (no matter if it was based on the wrong premise), to determine the run time for a WU, was serving me well for a couple of hundred of WUs.
It is then a big coincidence that EXACTLY this and only one WU does not 'fit' into the expected time frame of completion.
I would not even mention it, if we were talking about a difference of 12 hours or even a full day, but not having the slightest 'clue', how much longer this WU will last, I do not know, what to to.
Should I abort it, because I can not run it for much longer and if I understood correctly, this could be days or even weeks?!
In the first place, I'd crunch ONLY WUs, that last no more than 3 days, namely the 'cms short' apps. But since there are not any available at the moment, does that mean, that if I have time only for tasks, which last for 3 or maximum of 4 days, I can not participate in the project anymore?
In this case my assumption was wrong. But nevertheless the 'method' I described in my previous post (no matter if it was based on the wrong premise), to determine the run time for a WU, was serving me well for a couple of hundred of WUs.
It is then a big coincidence that EXACTLY this and only one WU does not 'fit' into the expected time frame of completion.
I would not even mention it, if we were talking about a difference of 12 hours or even a full day, but not having the slightest 'clue', how much longer this WU will last, I do not know, what to to.
Should I abort it, because I can not run it for much longer and if I understood correctly, this could be days or even weeks?!
In the first place, I'd crunch ONLY WUs, that last no more than 3 days, namely the 'cms short' apps. But since there are not any available at the moment, does that mean, that if I have time only for tasks, which last for 3 or maximum of 4 days, I can not participate in the project anymore?
-
Michael H.W. Weber
- Vereinsvorstand

- Beiträge: 23174
- Registriert: 07.01.2002 01:00
- Wohnort: Marpurk
Re: WU 6328245
There are indeed a few (!) exceptional WUs, that do not match your previously used runtime estimation, so in general it should be OK for you.
If you need to shut your system down, make sure you have the "keep app in memory" flag switched on, pause that WU and go to hibernation mode. In this case the work should not be lost and resume computation once you restart the system and resume that WU.
Michael.
If you need to shut your system down, make sure you have the "keep app in memory" flag switched on, pause that WU and go to hibernation mode. In this case the work should not be lost and resume computation once you restart the system and resume that WU.
Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.


-
Seeker2000
- Idle-Sammler

- Beiträge: 3
- Registriert: 07.12.2012 20:04
Re: WU 6328245
Thank You for your explanation. I understand (at least now I do), that the estimated run-time of a WU on the R.S. is only a 'guideline', and that in some cases the time estimate can be 'way off' of the 'real' time needed to complete the task.
But something I do not understand.
The WU in question has an estimate of around 460 000s on a PC, which is at least by its calculated speed (measured FPS, integer speed) roughly 2 or 3 times (at least) slower than mine.
Yet, I'm crunching this WU now for a little less than 500 000s and it seems, this number will only grow.
As you said:
The estimates can be in EXCEPTIONAL cases very different (longer), but is it even possible, that my computing time (on a PC that should be at least theoretically 2 or 3 times faster than the reference PC is) is actually VERY MUCH LONGER, than predicted by the 'estimate' OR is there another explanation possible (a bug in the WU,...)?
But what interests me most is, what should I do now with the WU, that is now running for almost 6 days and I have no clue when (or even if) it will finish.
Thank You again.
D
But something I do not understand.
The WU in question has an estimate of around 460 000s on a PC, which is at least by its calculated speed (measured FPS, integer speed) roughly 2 or 3 times (at least) slower than mine.
Yet, I'm crunching this WU now for a little less than 500 000s and it seems, this number will only grow.
As you said:
The estimates can be in EXCEPTIONAL cases very different (longer), but is it even possible, that my computing time (on a PC that should be at least theoretically 2 or 3 times faster than the reference PC is) is actually VERY MUCH LONGER, than predicted by the 'estimate' OR is there another explanation possible (a bug in the WU,...)?
But what interests me most is, what should I do now with the WU, that is now running for almost 6 days and I have no clue when (or even if) it will finish.
Thank You again.
D
-
yoyo
- Vereinsvorstand

- Beiträge: 8170
- Registriert: 17.12.2002 14:09
- Wohnort: Berlin
Re: WU 6328245
Let it run.
-
Michael H.W. Weber
- Vereinsvorstand

- Beiträge: 23174
- Registriert: 07.01.2002 01:00
- Wohnort: Marpurk
Re: WU 6328245
Yes, that isn't only possible, it happens quite frequently in the case of long WUs. An no, there are virtually no cases where the WU is buggy. This project suffers a bit from the fact that the runtime estimates cannot be accurate due to stochastic elements in the code.Seeker2000 hat geschrieben:The estimates can be in EXCEPTIONAL cases very different (longer), but is it even possible, that my computing time (on a PC that should be at least theoretically 2 or 3 times faster than the reference PC is) is actually VERY MUCH LONGER, than predicted by the 'estimate' OR is there another explanation possible (a bug in the WU,...)?
Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.

