I have an odd problem with one of my CM3n models. Since day one it has been running in high priority mode. This was caused by a wildly inflated initial to completion time of 6550 hours. The model is presently 61% complete (533 hours elapsed time). The to completion time is now down to 609 hours (350 would be more realistic) and the deadline is 11/23/11. That?s 70 days (1680 hours) till the deadline. With that kind of cushion, why is it still running in high priority mode?

---

It's a BOINC thing. When your box has run more of this kind of model it will predict things much more realistically. I don't understand the algorithm used, just know this to be the case.

Dave

I I have receive 6 long task with about 3000 hours to the deadline. Now it take about 20 hours by 1% so it will take 2000 hours to finish i.e. 83 day to finish 24 hours a day. The dead line is in 73 days ( 1er december ) what must I do ?

Your computer isn't all that fast, it appears to be a laptop, and you're running lots of other projects. Also, those type of climate models don't scale well when so many are run together.

So I'd suggest:
1) Set this project for No new tasks
2) Suspend all except one of the climate models and keep the rest for later
3) Wait and see if the time-to-completion drops after a week or so

You may be able to run a second one if BOINC isn't too overwhelmed by just one, but you really need to set your number of processors to 1 or 2 for future work.

And, as has been posted many times over the years, there is no real deadline.
But the researchers will be disappointed that they'll not get their results for years.

Thread by the researchers about the model: Welcome to the RAPIT experiment!.

---

Backups: Here

Hi thank you for your comment. It's strange because, yes I have a laptop but it's a core i7 QM720 1.6 Ghz with 8 CPU and I have a GPU. When I suspend other work, the WU use only 1 CPU so He don't run faster ! I just want to know if the process time will be discard if I go over the deadline time or Climate Prediction let the computer finish the task.

CPDN does not enforce the deadlines. It will allow you to finish. The results will be excepted and the data used whenever it is returned.

---

Thank you so much for the respond. I will continue with these WU.

This is great news
I have an atom cpu running only CPN models and it seems to take forever.
The deadline is approx 3 months but still it is not sufficient (according to calculations made with time allready run and percentage of completion)
So i'll let it run after the deadline.
Thanks
ccandido

---

Thanks for the info.
I started my first WU a week ago and it went into high priority when it started. Initial estimate of time was 611 hours. It's gone down as low as 580 hours, then I had a second WU appear, and ended-up running 2 WU high priority.
The first WU has climbed up and passed 590 hours, while the second WU started with 611 and now suggests 635 hours.

Reading through other threads where people mention problems with seg faults and error 193, I thought this was going to be another case of wasted efforts (been-there, done-that, with early versions of another BOINC project before they got cleaned-up a fair bit (leaving name of project out, since it's not important and the problem is fixed)).

Will take your advice of stopping the second climate prediction WU project for now and let the 2nd CPU run something else other than climate prediction until the 1st WU completes.

Out of curiosity, has anyone run multiple climate prediction models at the same time? Do they behave okay? (...just asking since I've run into some projects which used "static" variables, which may seem like a good idea at first, but isn't a good idea if you run more than one instance at a time).

Only problem running multiple instances of the same type of model is that it seems to slow them down, particularly with the full resolution ocean models. I don't notice it on the HADAM3P regional modelsThat doesn't mean the effect isn't there, just that I haven't noticed it. I haven't ever had errors that could be attributed to running two models of same type at same time.

Dave

Dave wrote:
>> Only problem running multiple instances of the same type of model is that it
>> seems to slow them down, particularly with the full resolution ocean models.
>> I don't notice it on the HADAM3P regional modelsThat doesn't mean the effect
>> isn't there, just that I haven't noticed it. I haven't ever had errors that >> could be attributed to running two models of same type at same time.

>> Dave

Just so. Been running 2 or 4 hadcm3 full resolution models on several machines for months now -- no problem.
As for slowing down with multiple same-type models it hasn't been much with 2 or 4 cores -- but my new 6-core machine I've only let it use 5 cores for these models because slowdown is likely 10% give or take. Saving one core for the OS and the sub-threads that fire up seems a good idea if only to save cache-trashing.
And to have one core free for whatever else I might want to run from time to time.

As for the evil error 193 -- who knows -- exclude boinc directory from virus scan, be sure to get clean suspend and exit before any backup or reboot -- I've tried several times to rerun these 193 failures from backups with no luck so far.

Cheers
e

---