Especially, In these large workunit, I think that the uselessness is remarkable.
Is there any Cancellation method?

---

I\'m not sure I understand your question, are you referring to the number of \'results\' issued per workunit?

If you look at the parameters for the workunit, you\'ll notice an entry called \"initial condition parameter\". My limited understanding of this is that it\'s a very small random factor applied to each of the individual results (i.e., the \'butterfly effect\'). So each result will be a little different, and the project looks at the range of climates generated (the group of slightly varying climates used in the forecast is called an \'ensemble\').

If you want to know more about this, google the words \'ensemble\' and \'climate model\'. The following link (PDF document) gives an outline:
http://www.cru.uea.ac.uk/~sarahk/pdf/es4.pdf

---

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

Hi Boris

For statistical analysis, the researchers in Oxford always need as large an ensemble of models as possible. The validity of the research method has been discussed in published articles.

The UK Meteorological Office also uses ensembles of (much shorter) models for predicting the weather. This has helped to increase the accuracy of their forecasts.

---

Cpdn news

Thank you for your reply.

I want to say that running same workunit with more than 2 computers is very loss.
Slower computer must report the results that was reported by faster computer already.
I thought that this problem is serious on Climateprediction.net project.
So, please tell me if reporting the results that was reported already is useless or not.

Sorry for my poor English.

---

Your English is excellent, Boris.

No, it isn\'t useless, it\'s useful because the programmers need duplicate complete copies of some results to compare and validate them. About 1.5% of models they compare are rejected, sometimes because the computers were overclocked and produced abnormal results for example with spikes in the graphs. The Oxford team wrote a research paper about the validity of the results our computers produce.

But mostly when the server sends 4 or 6 or 8 copies of a workunit to different computers, only one or sometimes two are completed. Look at this model which I\'m currently crunching. The server sent out 8 copies on 10 Aug.

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/workunit.php?wuid=6078707

Computer #1 in the list is me and I think I\'m processing it properly.
#2 is Zonker whose computer is downloading models but not processing them. We need to investigate this.
#3 crashed with a 22 error code, so the member probably did something that crashed the model, and didn\'t have a backup to restore and continue it.
#4 is Stolichnaya who sounds Russian and is processing the model properly.
#5 aborted the model.
#6 and #7 never started processing it.
#8 crashed the model with a 107 error, probably a graphics problem.

So of the 8 models Oxford will probably receive 1 or 2. Other models with no experienced crunchers on the list are less likely to produce 2 completed models.

Could you please tell us which of your models you\'re not happy about, and unhide (show) your computer/s so we can see your results?

---

Cpdn news

And even if multiple people do complete climate models from a single work unit, they are expected to be different if the \'Initial Condition Parameter\' is set. As I suggested in my earlier post, please search the net for \'ensemble\' and \'climate model\' if you want to know why they do this.

---

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

Boris

It looks like you\'re assuming that this climate project uses a quorum of several runs to validate a model, the way the other projects do.
Well it doesn\'t. Each model is slightly different in some small way, which is deliberate.
So each model is looked at separately, and without regard to whether or not others in the same batch have finished, have failed, or are still running.

---

Backups: Here

Thanks everyone.

I was able to understand that my thought(It is loss to run with more than 2 computers) was wrong.
Computation error is very serious in these project which receives a little number of results.
I think that our results is useful for verifying each other.

>>mo.v
I am pleased because you say \"Your English is excellent.\"
I am high-school student and studying English, so your message gained my self-confidence. Thank you.

---

I\'m a modern languages teacher in England so I easily recognise good students who have been in good schools with good languages teachers!

---

Cpdn news