Skip to content
mturilli edited this page Aug 31, 2013 · 2 revisions
  • Q: What is the input to troy as a whole?
    A: One or more workloads defined as a set of task descriptions and a relation among tasks. WorkLoad = {Task Descriptions}, R

  • Q: How much burden do we put on the application in order to use troy?
    A: We need a layer between Application and set of tasks. That is the layer where the application specific decomposition happens. E.g. Skeletons, Abstract Applications, Scientific Gateways.

  • Q: What would be a “good” workflow description language?
    A: Mark started on a WorkflowMgmtSystem for Pilots earlier, called Marvin, based on the GWENDIA language.

  • Q: What/where is the boundary between Workload and Workflow management?
    A: Workflow knows about application semantics; Workload knows about tasks syntax.

  • Q: What if not all tasks are available at T=0?
    A: Application workload can be decomposed into tasks at multiple stages in time. Therefore, a feedback loop back “up” is needed.

  • Q: What are the interfaces used by TROY?
    A: TROY-API: The interface to the TROY layer used by the application and/or frameworks. Pilot-API: The interface to the Pilot layer used by the Pilot Provisioner and Unit Dispatcher.

Clone this wiki locally