climateprediction.net home page

Posts by derekm

1) Message boards : Number crunching : 5.8.11 and error messages (Message 27089)
Posted 2795 days ago by derekm
Just upgraded to 5.8.11 and got this message:

25/2/2007 19:44:36|climateprediction.net|[file_xfer] Started download of file sulphur_data_4.22_windows_intelx86.zip
25/2/2007 19:44:47|climateprediction.net|[file_xfer] Temporarily failed download of sulphur_data_4.22_windows_intelx86.zip: file not found
25/2/2007 19:44:47|climateprediction.net|Backing off 1 hr 54 min 21 sec on download of file sulphur_data_4.22_windows_intelx86.zip

on and on...

Why is it happening? Project was (and still is) supended and not allowed to new tasks...


It\'s something to do with the upgrade to 5.8.x and not seeing the XML tables as the previous versions did. It will probably try to download the apps for all your projects. Don\'t worry, BOINC will honor the no new work or suspension. It is just trying to get the apps themselves.

If the app does not exist, as it may not for this older CPDN 4.x version, it may never download and eventually give up. It think that\'s what happened to me. I never had to reset it or anything else as drastic.
2) Message boards : Cafe CPDN : Cool Team Names / What do they mean? (Message 3554)
Posted 3695 days ago by derekm
> Hi all,
>
> I am impressed by the diversity of team names that are out there. I was
> wondering what some of them mean or how they came about, maybe some of their
> history with distributed computing or other online initiatives. Would any
> founders of the top teams like to comment?
>
> Everyone could pipe in on their vote for 'coolest' team
> name/purpose/philosophy... ;-)
>
> Mondoshawan @ 488°C
>
Click on the team name to go to the stats page for that team. The first two fields should be Description and Web Site. But I see you already knew that.

Then there is the Profile, but I see you have that covered too.

Beyond that, my suggestion is to go to the three other current BOINC projects and look for the same team to see how popular it is. SETI@Home is the oldest BOINC project and should give the best indication of a team's "history" in BOINCing. Of course SETI@Home and CPDN also have a long history of "classic" distributed computing teams.

derekm
Team MacAddict (vote for Max)
3) Questions and Answers : Preferences : Keep Work for xx Days not Working Correctly? (Message 3549)
Posted 3695 days ago by derekm
> when you first start the time-to-finish estimate will be way off, but after
> say 50-100 timesteps it should be a better estimate. It's about 3 weeks to
> finish a full CPDN workunit (3-phase or 45-year model run) on something like a
> Pentium IV 2.4GHz computer.

Since I started this thread, I'll try to finish it...

BOINC is doing a better job now that 3 of 4 projects are online with v4 (for Mac OS X that's the limit :-( no LHC). I am sure there is some complicated math that throws everything off a little when you start a project and then have another come online a week later at no credit yet. I can't say exactly how close my current time slicing is to my preferences, but we'll call it a work in progress that seems to be heading in the right direction. I predict(!) that when Predictor comes online and BOINC sees my 65+% allocation to it, I'll be a little slow with my CPDN timestep trickles for awhile.

One thing I have noticed is that if I have only CPDN and SETI being crunched at a pref of 10:1 on a dual proc machine, it still likes to preempt the CPDN and run two SETIs at once. You would think any good algorithm wouldn't do that (as shown in an example on the BOINC website). Oh well. I have more important things to worry about in life.
4) Questions and Answers : Macintosh : Multiple Projects (Message 2264)
Posted 3704 days ago by derekm
I assume from the context of your question that you have BOINC 4.05 and it is already an executable.

1) Open Terminal
2) Drag the BOINC app to the Terminal window.
3) Add "-attach_project" to the end of the line (the location of the BOINC app from step 2) making sure to have a space before "-attach_project" (don't type the quote marks)
4) Enter http://climateprediction.net and your Account ID key when prompted. You do have an Account ID, don't you? If not, you skipped the most important step... http://climateapps2.oucs.ox.ac.uk/cpdnboinc/create_account_form.php
5) Join Team MacAddict!!!!!

derekm
Team MacAddict
5) Questions and Answers : Preferences : Keep Work for xx Days not Working Correctly? (Message 2115)
Posted 3705 days ago by derekm

> Now that SETI have upgraded to v 4, I've today successfully downloaded a SETI
> work unit to go with my CPDN one. BOINC temporarily stopped my CPDN unit, and
> is currently chomping through a small SETI unit. Once it's finished, I'm
> presuming it'll go back to the CPDN unit again.
>

Ok, same for me too. It has gone back and forth between the two a few times. However, from what I can tell so far, the Resource Share setting is only for amount of work downloaded at once or in the cache. It has no effect on how often the suspensions take place. It also seems that BOINC tries to keep work in the cache for all projects at once. Therefore, all projects will eventually share CPU time evenly no matter the Resource Share setting unless the scheduler is not responding for a particular project. I really hope this is changed.

derekm
6) Questions and Answers : Preferences : Keep Work for xx Days not Working Correctly? (Message 1981)
Posted 3707 days ago by derekm
> One of the features of 4.xx is pre-emptive scheduling, which will allow BOINC
> to switch projects every hour. This means you will be able to do work for
> other projects while still working on the big WUs CPDN has. As soon as the
> other projects come online with the new CC, we shall see how well this works.
> :)

That must have been what I read somewhere. Thanks for the reminder. However, I do not see a preference for this under General or CPDN prefs on my account page. Is this behavior controllable? Will BOINC see the fact that there may be a lot of P@H units with short (1-2 day) deadlines and compensate by doing these more often to make sure the deadline is met and then switch to the others (no matter what I put in my individual project prefs for the Resource Share)?

Am I thinking too much about a problem that may not ever exist? Feel free to say I am. It wouldn't be the first time.

derekm
Team MacAddict
7) Questions and Answers : Preferences : Keep Work for xx Days not Working Correctly? (Message 1963)
Posted 3707 days ago by derekm
If my prefs are to keep 1-3 days (or something similar) of work, and given that a CPDN unit is allegedly going to take 800 more hours to process, why does BOINC insist on contacting P@H and S@H to get small bits of work (usually about 11000 seconds per request)? I obviously have a computer attached to all three projects at once.

I am aware of the problem before of BOINC ignoring home vs work vs school settings and using the defaults instead (fixed yet?). I am also aware that S@H and P@H are not compatible with v4. However, these are separate issues.

Does BOINC consider a CPDN work unit to be a trickle instead of the entire run? Is that the reason that BOINC insists on trying to get non-CPDN work when there is still a lot of CPDN work left to do?

I\'m afarid that BOINC will download a S@H/P@H unit but CPDN will crunch past the P@H/S@H return deadline. Either that or the P@H/S@H work will preempt the CPDN work because of the closer return deadline and my CPDN work will go undone for a year.

Didn\'t I read something about being able to share work across a single CPU so that multiple work units could be completed at the same time?

Sorry for the rambling post. Thanks in advance.

derekm
Team MacAddict



Copyright © 2002-2014 climateprediction.net