Grid jobs with large I/O

[Bender Rodriguez]

I have experienced that task results are persisted by the Digipede DB so that the DB fills up for jobs with large amount of output data (several GB). Also there is the penalty of an additional network hop because the data is directed through the digipede server.

For this reason I have been writing task results back to the client through direct remote I/O. I am wondering whether there is an efficient alternative, using the capabilities of the Digipede framework?

[Robert W. Anderson]

There is no facility on the Digipede Network to coordinate moving data in the way you need. I would be interested to hear what facilities you are using for such communication if you care to share what you are doing.