climateprediction.net home page
Longer Deadlines - That's What I'd Like To See

Longer Deadlines - That's What I'd Like To See

Questions and Answers : Wish list : Longer Deadlines - That's What I'd Like To See
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 47870 - Posted: 25 Dec 2013, 20:51:42 UTC

I seem to recall deadlines of almost a year at one time.

That was before this project changed it's direction, and became "pro".
These days, the work is created by professional climatologists, external to the University of Oxford, who pay the project to run big data sets for them.


This project relies on, and is aimed at, people who's computers are either dedicated to just this project, or give it most of the "spare clock time".
The more projects that people run, the less this project is suitable for them.

Others are, of course, welcome to contribute, but things are NOT going to be changed to cater for their many and varied requirements.

And, once again, there is NO deadline.
What WILL happen though, is that the model in question will not get the "Over Completed Success" messages. Rather it will be something about "Timed out - no response". Not quite accurate, but that's the best that the server software can do.

And then the server will re-issue it, in the hope that the next computer CAN complete it in time. (Mine take just on 3 weeks to crunch 1 to 4 of the long ones.)
And it will be the model that actually completes that is used to continue on to the next segment of what may be a VERY long total run, depending on what the researchers want.


ID: 47870 · Report as offensive     Reply Quote
Profile ex_brit
Avatar

Send message
Joined: 26 Aug 04
Posts: 84
Credit: 351,331
RAC: 0
Message 47871 - Posted: 26 Dec 2013, 1:12:30 UTC - in response to Message 47870.  

Normally mine would complete in that time but certain other projects have found out a way of circumventing BOINC and making every unit they put out high priority...will ditch same after I've had a fight with them.
ID: 47871 · Report as offensive     Reply Quote
Profile JIM

Send message
Joined: 31 Dec 07
Posts: 1152
Credit: 22,053,321
RAC: 4,417
Message 47872 - Posted: 26 Dec 2013, 2:21:39 UTC - in response to Message 47871.  

Normally mine would complete in that time but certain other projects have found out a way of circumventing BOINC and making every unit they put out high priority...will ditch same after I've had a fight with them.


What is your work buffer set at. I have found that some projects will run side by side with CPDN, sharing the processor time if the work buffer is set low (1 or 2 days), but, will immediately go to �high priority� and hog all processor time if the work buffer is set long, such as 7 to 10 days. I think the trigger may be if the buffer setting is greater than the other projects deadline.
ID: 47872 · Report as offensive     Reply Quote
Profile ex_brit
Avatar

Send message
Joined: 26 Aug 04
Posts: 84
Credit: 351,331
RAC: 0
Message 47874 - Posted: 26 Dec 2013, 11:29:44 UTC - in response to Message 47872.  
Last modified: 26 Dec 2013, 12:06:03 UTC

0.9 + 0.1 for a total of 1 day. There are a max of 8 WU's processing at once. (4 cores, hyperthreaded for a total of 8).

If I make the buffer higher BOINC loads far too many WU's from some projects.

In answer to the comment earlier about running too many projects. I have that arrangement simply because often projects run out of work, and at least this way I'm pretty-well guaranteed of getting something from somewhere.

Before someone comments, recent WU's have been a disaster for a number of reasons including the fact that I had a hardware failure and had to reinstall Windows. Thereby losing everything.
Also I did abort one or two because they had exceeded there deadline by quite a wide margin.

Someone earlier suggested limiting it to certain CPDN projects only, not sure if that would help.

I'm loathe to give up on such a great project but if needs must, I guess I'll have to.
ID: 47874 · Report as offensive     Reply Quote
Profile Iain Inglis
Volunteer moderator

Send message
Joined: 16 Jan 10
Posts: 1079
Credit: 6,906,534
RAC: 6,466
Message 47875 - Posted: 26 Dec 2013, 22:54:25 UTC

The project has been contemplating, and indeed testing, a variety of new models. Most of these new models are regional, reflecting the wider need to assess real-world impacts now that the general trajectory of climate change has been established.

You can through your preferences limit the involvement of your computing resources to only these regional models, which are shorter and have long deadlines relative to their run times. The HADAM3P regional variants are good BOINC citizens, as far as I know.

So there should be no need to give the entire project up. The longer HADCM3N models violate the "spare time" rule, as Les suggests, and many people will not run them for that reason.
ID: 47875 · Report as offensive     Reply Quote
Profile ex_brit
Avatar

Send message
Joined: 26 Aug 04
Posts: 84
Credit: 351,331
RAC: 0
Message 47877 - Posted: 26 Dec 2013, 23:23:59 UTC - in response to Message 47875.  

Thanks, I've already reset my preferences to that. Tine will tell of course.
ID: 47877 · Report as offensive     Reply Quote
Previous · 1 · 2

Questions and Answers : Wish list : Longer Deadlines - That's What I'd Like To See

©2024 climateprediction.net