cmsearch 0.13 run time

[Gary]

Hi.

I have a cmsearch 0.13 unit at 80 hrs. It's a methanococcoides-burtonii unit. The progress keeps bouncing back and forward between around 0 and 12% and it shows 151 hours still to run and has been on high priority for days. Is it OK?

Thanks.

[MonsieurPoot]

If you are talking about this WU of yours, the behavious is unusual:
http://www.rnaworld.de/rnaworld/workuni ... id=1126468

Your wingman needed around nine hours to finish the workunit and with your computer it should not take more than 15 hours. Do you have it running continuously? The cmsearch app does not make use of checkpoints, so every time you restart BOINC or the computer, the computation will begin anew from the beginning and all work that has been done so far is lost.

- Poot

[Gary]

Yeah, that's the one. It's currently at 81:11 hrs, 5.28% and showing no sign of stopping.
Yes, it's run continuously. That is to say that I haven't stopped it and the time hasn't reset. Just the progress. The lack of checkpoints is just unacceptable. I presume I'm going to get naff all if I cancel it, so what to do?
I will say that I've had no trouble with any of the many others I've crunched, though.

Cheers.

[Michael H.W. Weber]

I have a cmsearch 0.13 unit at 80 hrs. It's a methanococcoides-burtonii unit. The progress keeps bouncing back and forward between around 0 and 12% and it shows 151 hours still to run and has been on high priority for days. Is it OK?

This is a stadard WU that may run a bit longer than average. If it indeed is running 80 hrs already then, suprisingly, something faulty has happened. You should consider aborting it.

Michael.

[Gary]

I have a cmsearch 0.13 unit at 80 hrs. It's a methanococcoides-burtonii unit. The progress keeps bouncing back and forward between around 0 and 12% and it shows 151 hours still to run and has been on high priority for days. Is it OK?

This is a stadard WU that may run a bit longer than average. If it indeed is running 80 hrs already then, suprisingly, something faulty has happened. You should consider aborting it.

Michael.

Finally, I decided get rid of it. Look, I appreciate the work being done but WUs that don't checkpoint are just a joke. Please don't inflict this sort of nonsense on the crunching community. We deserve more respect than that. I hope you will at least have the decency to honour the amount of work lost and award the credits.

Thanks.

[Michael H.W. Weber]

I have a cmsearch 0.13 unit at 80 hrs. It's a methanococcoides-burtonii unit. The progress keeps bouncing back and forward between around 0 and 12% and it shows 151 hours still to run and has been on high priority for days. Is it OK?

This is a stadard WU that may run a bit longer than average. If it indeed is running 80 hrs already then, suprisingly, something faulty has happened. You should consider aborting it.

Michael.

Finally, I decided get rid of it. Look, I appreciate the work being done but WUs that don't checkpoint are just a joke. Please don't inflict this sort of nonsense on the crunching community. We deserve more respect than that.

We have actually explained it in advance several times, here is the our most recent comment on this issue.

Michael.

[Gary]

Hi Michael,

We have actually explained it in advance several times, here is the our most recent comment on this issue.

Again, with due respect for you and the work, that's neither here nor there. Firstly, who has the time or inclination to troll through yet another of the dozens of forums we have to deal with on a daily basis? Secondly, that's just the point. If you want to take advantage of the power we offer your obligation is to provide work that meets the system's requirements, not do the minimum necessary to serve your own purposes at our expense. The assertion in that post that providing checkpointing is "technically impossible" is rubbish. You just don't want to do the work. I understand the reasons for that but that is just it, isn't it? If you want to benefit you have an obligation to do the right thing by us.

Cheers.

[mxplm]

Hey Gary,

Firstly, who has the time or inclination to troll through yet another of the dozens of forums we have to deal with on a daily basis?

Nobody, and nobody expects you to do so. Since it's a question concerning the Project, you could have taken a look at the RNA World Frequently Asked Questions.

The assertion in that post that providing checkpointing is "technically impossible" is rubbish. You just don't want to do the work.

As Michael explained: to support native checkpointing (done by the science app itself), the whole program would have to be rewritten from scratch. This of course is not impossible, but it's close to that for us. And this is why we are looking for another solution, e.g. the VM approach, which Michael also mentioned in his post and in the FAQ. I hope you understand that we work on solving the problem, we just didn't succeed by now. Please also keep in mind that RNA World still is in Beta status.

[Ananas]

We actually might have something "checkpoint-alike" before the VM solution comes along :


cmsearch :

The workunits can be split, crunched in parts and then reassembled, which will solve memory and runtime problems partially. Each of those "chunks" will give us one checkpoint.


cmcalibrate :

The workunits can be crunched in two parts, it will not save too much memory (unless the developer of the software cmcalibrate adopts our changes) but give us at least one checkpoint.

[FalconFly]

IMHO several workunits are still way too unstable for mass production. Add to that the lack of checkpoints and you're asking for a whole lot of frustration (unprepared users) and resulting troubles.

I just wasted another ~500 hours of cpu time over hung workunits I'd rather had invested in something more useful than code bugs

-- edit --

*ugh* They finally finished, after displaying 100% complete for many hours.

http://www.rnaworld.de/rnaworld/result. ... id=2842245
http://www.rnaworld.de/rnaworld/result. ... id=2842252
http://www.rnaworld.de/rnaworld/result. ... id=2842259
http://www.rnaworld.de/rnaworld/result. ... id=2842279
http://www.rnaworld.de/rnaworld/result. ... id=2842521

[apofis]

Draft RNA World does not store intermediate data of their calculations.
Got the job 22 hours, turn off the electricity when it was performed on 90%. Turn on the computer, but it started counting from 0%.
This happens when the BOINC press pause, and then launch. Or stop the program. (With priostonovkoy assignment). At Rosseta @ Home and GPUGRID No more.

No intermediate records, it is impossible to perform long tasks.

[mxplm]

That's correct, by now there is no checkpointing in cmsearch. See the FAQ

Long tasks sometimes take some tries before they get done. If you activate the option "Leave app in memory while paused", you can resume the task after pausing without restarting from 0%.