Hiya, I\'ve just started one of the simulations going and am keen to participate (let me make this absolutely clear) as it\'s an issue that I care about a great deal. But to be brutally honest, I\'m a bit sceptical about the design of the experiment (I\'m sure that someone else must have made these comments elsewhere, but it\'s not easy to search this forum for what I want to say/ask, so I\'ll say it myself).

The point is that other distributed computation projects like SETI@HOME do their work in much more discrete, bite-sized chunks, whereas this experiment (according to what I see on my BOINC screen, and if I\'ve interpreted correctly) is going to run for another 4832 hours. The deadline for the results from my computer is November 18th 2007. I have also put SETI@HOME and Einstein@HOME on my BOINC thingy, meaning that the climate change experiment has 33.3etc% of the resource share. And on the new 5.0 version of BOINC I don\'t see where I can change these figures, to give the climate simulation 60%, say, of the resource share (why on earth they would have taken that feature away I don\'t know).

Now, let\'s do the math... How many days til 18/11/07? 510/520? Roughly. How many days is 4832 hours? 201 and a bit. What\'s one third of 510? 170. So even if I leave my computer on 24-7, I won\'t make it in time. And even if I weren\'t dividing the processing along with SETI and Einstein, my computer would have to be on an awful lot more of the time than I guess most home computers generally are.

Factor in the fact that the average Windows user probably ends up reinstalling the system at least once in a while. So let\'s say the simulation running on computer A is 95% done, and computer A fails completely, such that the only solution is format and reinstall. What happens? Is all the data lost? (I don\'t know, because more than half of the links I\'ve tried on this site don\'t work.) That would be a bit of a shame, wouldn\'t it? Or does it update the server as it goes along?

Last thing: the disk space that this simulation takes up on my machine is more than 700MB. Now, personally i really don\'t mind, and my PC does have room for that, but there must be plenty of well-intentioned people with 40 Gig disks who would prefer this thing to be a tad bit smaller.

Does anybody agree with this Or has something very strange happened on my machine? Is my version of BOINC a dud? As far as I know it\'s the latest one - 5.2.13.

I\'ll just say it one more time - of all the distributed computation projects, this is the one that I\'m most sympathetic towards, but I really think that, unless something is dramatically wrong on my PC, you should change the way it works.

Sorry to sound bolshie. I don\'t want to make trouble. I want to help, actually :~)

H

Never will change, you need to read up on what, who, when, where, just look at the column of info at the left....


Opps they did change, they have a seasonal project, takes about 20 days on my 840ee 3.2 3gig ram....

---

BOINC Wiki

As Jim said, more or less: Been there; Heard that.

This is the Olympics of DC projects, whereas the rest are just a jog around the park.
The other projects provide data for people to search through, looking for values that match a certain list of criteria.
This project does the opposite: it provides a small collection of values for feeding into the program, which then slowly creates a 3 Dimensional model of the Earth\'s atmosphere, at half hour intervals, for many years. This takes time.

There are / have been a lot of people who have felt like you: they came here to add this project to their list, found out what it involves, and left again.
And still we plod on, creating model after model.

The shorter, simpler \'slab\' models have provided a large range of vaild parameters, the sulphur models have provided the \'reasonable\' levels of sulphates for feeding into the TCMs, and the spinups have provided the starting values for TCMs.
But now we have finally arrived at what the project was designed to do: look into possible futures.

Climate modelling by it\'s very nature is a huge undertaking; even the Met Office supercomputers, which provided our basic programs, take well over a day to run a climate model. (Which is NOT the same as their usual weather models.)

Sorry, but that\'s how it is.

As for your hard disk space, that will be the left over from your previous models.
Slab and sulphur both left some data behind for the modelers to study, but because of complaints from people who didn\'t care about the science, this has changed. Now the models clean up after themselves.
Unless they crash, in which case the folders still need to be removed manually.
The TCMs should only get to about 600 Megs.

And on the new 5.0 version of BOINC I don\'t see where I can change these figures, to give the climate simulation 60%, say, of the resource share (why on earth they would have taken that feature away I don\'t know).

You have to go to your account on each project (e.g. [url=http://climateapps2.oucs.ox.ac.uk/cpdnboinc/home.php] for CPDN) and set the quota you want to assign to each project.

I have 300 for CPDN = 75%
I have 100 for S@H = 25%

This way I can finish a CPDN WU within a few months.

P.S. The \'To completion\' hours are very badly calculated, especially during the early stages of a WU. You get a better estimate if you take the processed CPU time and extrapolate it to 100%. The \'To completion\' hours will get better as the WU progresses.

OK, Guys, thanks for your help.

I think I understand a little better now, and will persevere. Like I said, the disk space was never a problem for me - I was just thinking that some people might get put off by that. Also, all the possible things that can go wrong with someone\'s PC during the time that the experiment is running... but fair enough, you take the risk, I suppose.

One thing I do have to say, though, is that I just had this error message:

\"Visual Fortran runtime error\"

which seems to be related to this project according to what my Google search dug up:

http://www.climateprediction.net/board/viewtopic.php?t=2296

Could anyone put all that into plain English for me? It\'s strange that that posting is from quite a while ago. I\'d have thought they\'d have sorted that out by now, with this latest version.

The immediate effect of this error for me seems to have been to mess up the way I have my Windows taskbar arranged. It\'s OK after restarting, but it was a bit irritating at the time.

OK, thanks again,

Chris

---

There are very many different \'fortran error\' messages, and i\'ve never seen that \'namelist\' error before.

Take a look at the topmost post in this forum for a list of suggestions to help avoid this type of error.

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Also, all the possible things that can go wrong with someone\'s PC during the time that the experiment is running...

Yes, things can go wrong, from power failures to Windows crashes. When CPDN is in the middle of multiple I/O operations when this happens, the project files may get out of synch, and the project will possibly crash upon restart.

To avoid this it is extremely helpful to have a restart point -> a backup.

I suspend (Commands, Suspend) my project(s) about once a week, close BOINC Manager, and take a full backup of the BOINC folder. If something should happen, I can always restore and take it from there.