Calculations of my job started from 2012 (in show graphic), now I've done 30% and am in 2013 =)
Why calculate that we already known? I can not understand.

As there is one problem - if you turn off the BOINC client, then there is a small pullback of about 30 minutes. How to avoid it? In client settings set to make checkpoint every 60sec (default).

I believe that the reason for modeling the past is to see if global warming is making extreme weather event more frequent. I think 2 sets of models are run, one with present enhanced CO2 levels and the other historic non-enhanced levels. This helps determine if recent extreme events like the recent floods in Europe are natural events or likely caused by human actions.

The reason for the �pullback� is that as the models run they set �save points� along the way. When the model is stopped and restarted it restarts from the most recent save point. You can see these in the graphic display. They only not to loose time would be to wait until you are just past a save point before stopping.

---

Welcome to the project and the boards, Tristis.

JIM's suggestion is easy if running one or two CPDN tasks, more difficult and time-consuming if you run eight CPDN tasks on your AMD processor. If one accepts statistics, over time, rerun time should "average out" to near the middle of the checkpoint interval. If one waits for each task to reach a checkpoint before stopping it, there will be stopped tasks and lost processing time while waiting for others to reach checkpoint. It's a coin-toss as to which method to choose.

CPDN isn't the only project with checkpoints -- how many others allow management in line with checkpoints?

CPDN checkpoints intervals vary among the several experiments.

Default every 60 seconds? Hardly! CPDN checkpoints are based upon logical processing points. A large number of files is open and have to be written to disk. Waiting for synchronization every minute would add to task's wall-clock duration.

I prefer to concentrate on "small" in your "small pullback" comment. How to avoid it? My solution: machines are shut down only when necessary, typically when software updates require it.

EDIT: There are links on the Main Page (top of list on left) to explanations of what underlies each experiment.

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

thanks.
yes i can shut down PC only when necessary, but the job will last for 300 hours, and it happens more than once)
I have many different options why you need calculate from 2012, but it seems to me better to give everyone one pre-calculated initial data, because they are really the same for all.
I know that somewhere in the depths of the site it's probably explained, but it's difficult to me read so many on english =)

Hello Tristis

The models are NOT all the same. Each starts with slightly different values for the many variables, and continues from there.
Each "model" is only a short part of a very long model that's put together from these short models. So where one finishes, (if it does!), the next one continues.

And they will only take that long if your computer is also running other projects, or only running part time.

**********

Each of the research groups around the world are working on different aspects of climate, some of which is about past events.
Some in Australia & New Zealand are working working on what may (or may not), have happened in Australia to cause recent heat waves and drought.

Some in Europe and the UK are work on the recent flooding in parts of the UK.
This type of research is known as "attribution".

A group in several universities in the UK and Europe are working on finding a set of model versions that can shadow recent surface and subsurface ocean changes. Once this model set is identified the range of uncertainty in the Atlantic Meridional Overturning Circulation (AMOC) for these model versions will be assessed.

Tristis,

Don't worry about your use of English, you do very well.

Historical note:
We started with a more-capable bulletin board. (Unfortunately, it was often attacked. [Back then, a primary moderators' task was eliminating trash posts.] Finally, the board so seriously damaged that it was taken offline.) One feature was effort to recruit linguists to translate posts to allow as many languages as possible to participate. There were several languages available but I don't recall whether Russian was among them. I wish we had that facility here.

Have you tried any of the Internet translation programs? (I've read that they make some laughable mistakes but also give a reasonably good "first look." With your facility for English, I'm confident you could spot the errors.)

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

So some models can't be generated, while previous not complete? it will be very long experiment. Pity that you can not use GPU.

In fact, such projects do not get enough attention. I accidentally found out about BOINC. If the results of advances published not only at specialized resources, it would be more volunteers.

astroWX
I can read all this info, but not quickly =)

The very last file that's returned, zip 13, is actually full of data that's needed to continue the work into the next section. So a computer has to get all the way through, to get to that point.

A new user, I have had several climate prediction tasks running for ten days or so then suspended without explanation. No task completed. Another project (home cosmology), started while waiting for cp tasks, steams ahead accumulating multiple 'credits'. My interest is in the climate prediction programs. Do I need to reset preferences?

---

Hi GH

Not sure what you're asking here.

suspended without explanation

All of your tasks have either completed, or failed. See the Tasks page here.

*************

Is it about no credits?
cpdn does the credits calcs differently to all other projects. And there been problems with this over the past year, as discussed in a thread in number crunching. The scripts now get run once or twice a month.

Thanks. The tasks page looks impressive. Prior to seeing this I suspected that jobs were failing possibly due to incorrect settings on my part. Can users access their task page routinely? I may have missed some advice about this.

---

Yes, that page is all part of what a user can see.

A bit more:
There's two types of Account pages:
1) The "public" page, where others can see SOME of what's on other people's page.
2) The "private" page, which has additional info which is only for the owner of the page.

There's several ways to get to the user's Account page:
1) A button in the Projects tab of your BOINC manager, labelled Your account. This goes to the public view.
2) Clicking on someone's name to the left of their post on these boards. This also goes to the public view, and is a quick way to check the errors on a poster's completed tasks to see the errors. Unless their computers are hidden.
3) Click on Your account in the blue menu to the left of here. This one goes to the private view.

Try MY name to see the public view, and then the blue menu to see the extra in YOUR private page.

***************

And you can go further.

Click on a model name in the left column of a Tasks list.
Scroll down to Stderr and click on the + symbol.
Now you get the expanded list of all the errors etc for the model.

***************

I have the Tasks page for both of my computers bookmarked into the Bookmarks Toolbar in my browser, so that I can do a quick check when needed.

This is just a quickie from a noob - the faq links to an article about backing up in a piece about 'computation error' - but it links to an article that has been blocked by moderators. My first data-batch hit this after less than 5 mins into a 2+ day routine - a little disconcerting when it's your first batch.

So:
i] is it worth putting a different, working link into the faq? I've yet to dig for 'how to back up' [I've never had to do it for seti@home, so this is a new one for me].
ii] is there a way of prioritising smaller batches for users with old and slow computers like me [often run by old and slow people, also like me] so we don't get allocated the humungous ones? I don't mind three days, but...

Cheers - F.

Hi Francis,
I suspect the link may go to the old PHP message board which was taken down because of the amount of spam on it and other problems.

The bit in the FAQ about backing up is rather out of date. Now that the majority of computers are multi-core, backing up is likely to take all other tasks back to the point of the last backup. restoring a single task would require a detailed knowledge of files such as client-state.xml and skilful editing of these files and generally is thought of as not worth it these days. Apart from tasks which crash due to anti-virus intervention and a few other causes, the vast majority of crashes are down to the model/task in question and will likely crash again at the same point anyway.

However I do agree that the FAQ needs updating. It is worth posting in here about problems with computation error if you want to be able to work out whether it is common in the batch that you experience it or whether you do need to look at your own setup.

Looking at the three tasks that have failed on your machine, the segmentation violation is when something in the program tries to use a memory area that it is not allowed to. This is down to the downloaded files for the task and has been widely experienced on a number of the recent batches. One of the forum moderators has sent an email to the project people about it but action is unlikely until Monday. Hope this helps.

See the thread on, "New Work" in the number crunching section for more detail on current problems.