Is anybody running SETI and climateprediction under BOINC 4.05

SETI has gone live on BOINC 4.05 as soon as it download the new client process and WU's climateprediction displayed the following.

climateprediction.net - 2004-08-29 20:08:15 - Preempting result 008b_300025284_0 (quit)

Any body else had the same problem??

Thanks in advance.

Kevin....

well... not sure what my error message was , but yes , mine went down on climate too. then i updated to it and it downloaded something real quick from that site and started crunching again.

Pre-empting is a new feature in v4, meaning every hour the client will, based on your resource-share, choose to either continue running current project or switch to another project.
Of course if it can't get work from one project it can't switch to this either. ;)

Note, if you haven't enabled "Leave project applications in memory while activities are suspended?" in your preferences, pre-empting will exit the current application then pre-empted.

There's still some fine-polishing to be done with the pre-empting, like making sure deadlines is kept & only quit right after a checkpoint...

Yes I have just had the same thing show up, tried to update like Dave but didn't work (yet), thanks for your info Ingleside, will try that in prefs.

Regards from Bonnie Scotland

Tried to update prefs like Ingleside suggested, project still not running, Seti is fine though, any suggestions folks ?

Regards from Bonnie Scotland

I think what Ingleside was saying is that you've told BOINC to divide time between CPDN and SETI, so it's through with CPDN for a little while while it devotes some time to SETI, then back to CPDN....

>>There's still some fine-polishing to be done with the pre-empting, like making sure deadlines is kept & only quit right after a checkpoint...

Are you saying that BOINC4 will ditch any work done after a checkpoint if it chooses to switch to another app? How frequent are checkpoints, or is that application dependent? With checkpoints at, say, 90 mins, and pre-emption at 60 mins it seems unlikely that any work would get done.

Maybe I'll just wait for 4.06.

Giskard - the first telepathic robot.

> I think what Ingleside was saying is that you've told BOINC to divide time
> between CPDN and SETI, so it's through with CPDN for a little while while it
> devotes some time to SETI, then back to CPDN....
>
>

Yes, if you've got 2 projects with equal share, example seti@home & CPDN, you'll crunch 1 hour CPDN, 1 hour seti@home, 1 hour CPDN, 1 hour seti@home...

With 33/66-share, 100 for seti & 200 for CPDN, you'll crunch 1 hour seti, 2 hour CPDN, 1 hour seti, 2 hour CPDN...

Not sure if it takes into consideration you maybe not having gotten any work from seti@home the last couple of weeks, so gives this a bigger boost to "catch-up" with CPDN, but basically it's this way it should work.

There's still some quirks to fix, atleast one surfacting in CPDN-beta was if dual-cpu & less share than 50% for CPDN, you only got work for one cpu, even no other project at the time could give you work...

>
> Are you saying that BOINC4 will ditch any work done after a checkpoint if it
> chooses to switch to another app? How frequent are checkpoints, or is that
> application dependent? With checkpoints at, say, 90 mins, and pre-emption at
> 60 mins it seems unlikely that any work would get done.
>
> Maybe I'll just wait for 4.06.
>
> Giskard - the first telepathic robot.
>

Yes, if you haven't gone into your preferences and enabled "Leave project applications in memory while activities are suspended?" the switching between projects will lose anything done between checkpoints.

As for how often checkpoints, BOINC has the "write to disk", but projects like CPDN doesn't use this functionality so they'll checkpoint then they wants and not then BOINC wants... CPDN checkpoints every 144 timesteps, with the fastest machines using 2 s/TS this gives every 5 minutes, while more "normal" machines is 10-15 minutes between checkpoints. Old p2/p3 is not fast enough, so will never checkpoint if you don't let the application use your virtual memory instead of terminating then pre-empted.

BTW, if the re-benchmarking works as it should now and re-benchmarks every 5 days, this also will currently terminate the process even you're only running one project...

So atleast for now, I'll recommend to enable "Leave project applications in memory while activities are suspended?"