Have any of work units helped your (project)statistics at all. Upon looking at my own results, I have noticed that out of the 40 plus work units that I have worked on over the last few years, not one was completed fully successfully (all client errors). I do know that some wu’s did complete more than half on some of them before screwing up. I recall that the older large ones (2500 hours) had 160 years and now the smaller ones (550 hours) have 45 year cycles. With all these client errors, was any useful information obtained?

I know that this is an unusual question for the group. But of the 14 billion seconds (over 167 work days; total credit 58k) on my computers helped at all? I looked over other computers results and notice that they have not been that many successful completions at all. Thank you in advance.
Fred Moldt
Siifred


Seconds 14,514,983.24 Credits 58167.18
Hours 4031.939789
Days 167.9974912

---

SIIFRED

Hi,

The climate models mostly upload their progress as they go, so an error isn\'t necessarily a big problem.

HadCM3 - Coupled models - climate summary uploaded after each model decade, and a full \'restart dump\' upload at 1960, 2000, 2040 and 2080.

HadSM3 - Slab models - climate is uploaded after the end of each phase (33%, 66%, 100%)

HadAM3 - SAP models - Climate uploaded at the end only (100%)

But since it\'s always more efficient (and satisfying) to complete models yourself, then I\'d suggest taking a look through the \'READMEs\' for tips to avoid crashes, and in particular the \'crashes and other problems\' one. The link is in my signature.

---

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

By using all the advice in the ReadMes and elsewhere, helpfully supplied by other crunchers, I have succeeded in completing all of my models. I think the total is 5 now (in the early days, my only PC was very slow and it took about 18 months to finish one model). But at some point, each one of them crashed at least once. I was able to restore them from backups and resume processing to take them to full completion.

Hi Siifred

I\'ve looked through your model results to see if I could find any specific things you need to pay particular attention to in the READMEs. There are a few.

Here\'s your list of computers. Click All hosts to see the complete list:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php?userid=131297&show_all=1&sort=rpc_time

If you click on computer #1 you find this list of models

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/results.php?hostid=338577

If you then click on each result in turn you find why each model crashed. On this computer the models all crashed with exit code 1. This most probably means you shut down the computer without exiting from boinc first. You need to select File - Exit in the boinc manager, or right-click on the boinc icon and select Exit. You then have to wait till the icon disappears before beginning the shutdown process, otherwise Windows shuts down before boinc does; this can ruin files.

Exit code 1 could possibly also mean that this computer\'s graphics card drivers need to be updated.

On computer #4 in the list the models crashed with a 107 exit code. This indicates graphics errors. I\'d advise you to disable the screensaver if you haven\'t already done so. It would be a good idea to suspend the model on this computer before you do anything graphics-intensive, like playing games. You should update the graphics card driver for this computer. (You don\'t pay for this - it\'s a free update from the web.)

On computer #5 there were some 107 exit codes, but you also aborted several models. Maybe you had too many models and aborted some. You can avoid getting models you don\'t want by going to the boinc manager Projects tab, highlighting cpdn and clicking the No new work/tasks button. The day you do need a new model, click the button again to allow new work.

In the README about avoiding model crashes, I\'d particularly recommend item #5 by Mike and item #6 by Thyme Lawn who explains how update graphics card drivers.

http://www.climateprediction.net/board/viewtopic.php?t=5896

In the same README, item #1 by Les explains an easy backup and restore method so that if your model does crash, you can restore the backup and continue the same model. There\'s a whole README giving other more sophisticated backup methods, but I just use Les\'s easy instructions.

If everybody who suffers model crashes posted on the forum as you have done to ask about it, the completion rate would be higher. Because the models are so long, the probability of something unfortunate happening before the end of a model is quite high, which is why the precautions really do help.

Hope these ideas are useful.

---

Cpdn news