Application:
HadCM3 Coupled Model Experiment Optimised File I/O 5.40
Name:
hadcm3inct_cnwz_1920_160_0589172_4
CPU time:
2417:14:43
Progress:
36.091%
To completion:
2652.34.00
Deadline:
20/04/2008


The time to completeion time seems to fluctuate by around 30 seconds but does not really change much.

Dual AMD 1.2 ghz processors.
31/10/2007 16:22:02|| Number of CPUs: 2
31/10/2007 16:22:02|| 1119 floating point MIPS (Whetstone) per CPU
31/10/2007 16:22:02|| 1842 integer MIPS (Dhrystone) per CPU


PS it would be useful if BOINC would allow the following:
a) copy and paste of a line from the tasks list.
b) have a copy and paste screen for the computer details and bench marks...

I seem to remember that originally the time to completeion was around the 2700- 3000 mark when I got the workunit in the first place.

this is the work unit:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/workunit.php?wuid=6040455

I have yet to do a functional workunit from CP - this particular work unit seems to have ended in error for every one else doing it, in the WU details is says:

minimum quorum 1
initial replication 4
max # of error/total/success results 3, 4, 1
errors Too many total results


is there a possibility that I can complete this work unit successfully?

---

Two of the three models that have crashed in that work unit never got started; the third crashed with a problem with the PC. So there\'s every chance that the model you\'re running will complete successfully.

Things do sometimes happen on the PC that cause the model to crash, so it\'s a good idea to take a backup from time to time. There is a description of how to do that in this post.

Best of luck.

a) copy and paste of a line from the tasks list.
b) have a copy and paste screen for the computer details and bench marks...

Changes to BOINC should be posted on the BOINC site. There\'s a wish list there.

b) There\'s no need for this; people can see most of the info needed in order to give advice just by clicking on your name to the left of your post.

Your computer is rather slow by todays standards, and you\'re running a lot of projects. (See - we CAN see this. :) )
It\'s possible that the resource share for cpdn is low and the models aren\'t running for long enpough to reach a checkpoint, and so the same data is being re-calculated each time the climate model is restarted.
You can see when a checkpoint has been reached by looking at a countdown timer in the graphics display - Z to relay the left hand overlay if it\'s there, and then 8 to show the timer. When it returns from zero to 432, the data should get stored to the hd. It will take a few seconds, so wait until it reaches 430.

You need to know how long it takes to make one full cycle through the timer, then set your \"Switch projects\" value to a bit more than this.

Setting the option to leave the model in memory is another way to avoid the re-calcs. The work done so far will then be saved when tasks are switched, instead of being dumped.

Re: the list in your second post
You need to be well informed to be able to interpret these. Too many total results for instance, isn\'t used by this project. Other projects use it to tell when the number of results returned exceeds the quorum.
Here the number of trickles returned ends up being counted, so it\'s obvious that this will far exceed any quorum.
And initial replication 4 is the number of models issued with the \'basic\' set of parameter values, but each of these also has another value, which causes each model to be slightly different.
A bit like the fingers on your hand; they\'re all on the same hand, but they\'re all different. Unless you\'re one of the aliens that the SETI people are looking for. In which case, for heavens sake give then a call! :)

The best way to ensure that you have the best possible chance of completing a model, is to make regular backups. See my sig for different ways to do this.
The notes on my copy and paste method includes a link to how to deal with mutliple projects.

---

Backups: Here

Good point re the ability to see my vital stats as it were on the website.

I run all the various projects I am signed up for since not all have work, and some I have retired eg seti/primegrid since they are not what I want to do now.

12 projects in all:
six are ot of work
four have work (inc CP)
2 are disabled

so CP does get deferred some of the time.


I have set my comp to use shorter CP WUs now in the settings so the next ones will not be notionaly vexatios to me.

My comp does crash now and then - but backing up the work is currently of no interest to me - although I do see the strong argument for the need for it in the case of CP especially in the case of these long WUs.

I will very definitly look into changing the \'switch projects\' timer from 120minutes to a new optimal value, going from your suggestions. This is probably the answer to the issue.

A side point is the data on old WU\'s in CP, looking through the older ones I participated on it seems the data about them has been reduced somewhat - so that it looks like some data has been lost.


Many thanks for your responses and help.

---

Two of the three models that have crashed in that work unit never got started; the third crashed with a problem with the PC. So there\'s every chance that the model you\'re running will complete successfully.

Things do sometimes happen on the PC that cause the model to crash, so it\'s a good idea to take a backup from time to time. There is a description of how to do that in this post.

Best of luck.

Frequent backups really do pay off. I learned this the hard way. Before I got into the habit of making daily backups I lost several models in various states of completion. Before I saw the light (about making frequent backups) I never got beyond 30% or 40%. My models still crash, but, now I just restore and go on. I had to do a restore just yesterday. Somehow my model got crashed while I was installing an update from Microsoft. I don’t know why, I had exited before I started, but, after the restart the model was crashed. A simple restore turned a potential disaster in a minor glitch. All I lost was 20 hours of CPU time. My model is now at 87.5% and crunching happily.