Posts by OBrash

1) Message boards : Server programs : Checkpointing when using wrapper (Message 97126)
Posted 27 Mar 2020 by OBrash
Post:
Thanks! I was not even aware of those. I have continued poking around and now have a little more insight as to how it is done, but still have some questions so I will post those there.
2) Message boards : Server programs : Checkpointing when using wrapper (Message 97122)
Posted 27 Mar 2020 by OBrash
Post:
I have been successfully using Boinc internally to distribute to a pool of machines.

I used the wrapper for the app and that has been working just fine. As the jobs get more complex, however, they are taking many days. I would like to implement checkpointing and it is not clear to me in the documentation whether that is even possible when using the wrapper instead of the API.

Could someone point me in the right direction.

Thanks in advance.
3) Message boards : Server programs : Terminating jobs early (Message 90706)
Posted 16 Mar 2019 by OBrash
Post:
Well, that makes me feel silly. One follow on questions: is there any bad side effect to calling cancel on workunits that have already been assimilated or do I need to take care and only call it on those that have not been returned yet?
4) Message boards : Server programs : Frequency of getting new tasks on local grid (Message 90705)
Posted 16 Mar 2019 by OBrash
Post:
Thank you. That is exactly what I was looking for.
5) Message boards : Server programs : Terminating jobs early (Message 90702)
Posted 16 Mar 2019 by OBrash
Post:
We have just start experimenting using Boinc to run a local grid for one specific application which involves a numeric search for a solution. If an appropriate solution is found, it might mean that no further tasks need to be executed for that job.

What is the recommended way for the assimilator to cancel all remaining tasks for that job if it finds a solution that eliminates the need for further searching? And, is there a way to notify the clients to cancel any tasks they might have queued up?
6) Message boards : Server programs : Frequency of getting new tasks on local grid (Message 90701)
Posted 16 Mar 2019 by OBrash
Post:
We are experimenting with using Boinc to operate an internal grid for one specific application. I have successfully configured a server using the wrapper with our application and our initial tests using a grid of about 20 scattered machines have gone quite well.

As we run tests, and the tests complete, the clients on those 20 machines sit idle. When we generate a new job on the server, it takes some of the clients many hours (in one case over 24) to actually get new tasks. Of course, if I hit the update button on the client it gets new work right away and starts computing.

Is there a configuration where I could tell the clients to check for new work more frequently? I would like it to be on the order of minutes rather than hours. I understand why it needs to be the way it is in a global volunteer environment but for this use case I am hoping there is an easy way to override it without having to dig into/modify the source.




Copyright © 2021 University of California. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.