Implement a Hive engine which can be executed in the Hive

[ascheibe]

It should be possible to send algorithms which use the Hive engine to the Hive. The Hive engine must be able to run on a slave and send tasks to the Hive. It should also be possible to pause algorithms which wait for child tasks to finish. They should be sent back to the server and started again when the child tasks have finished.

[ascheibe]

r7284 added branch for the HiveEngine which can be executed in the Hive

[ascheibe]

r7285

• added interface for the hive engine so that the parent task can be set which is needed for creating child tasks on the slave
• fixed uploading of the configuration file when executed on the slave

[ascheibe]

r7287

• added the new Hive engine. The engine can now be executed in the Hive. If child tasks are created it pauses and is transfered back to the server. If the child tasks are finished it is sent back to a slave.
• changed the server so that it reschedules paused parent tasks if their childs are finished as well as tasks where FinishWhenChildJobsFinished is set to false

[ascheibe]

r7320 fixed a bug in the job scheduler for hive engine tasks

[ascheibe]

r7401 updated branch from trunk

[ascheibe]

r7417 added an option so that the parent task can also wait on the slave for the child tasks to finish

[gkronber]

Please accept this ticket.

[ascheibe]

r13432 deleted Hive engine branch

[ascheibe]

I'm terminating this branch and ticket as it did not work out as hoped. A better approach for fine-grained distribution of work is needed.