70 likes | 80 Views
This paper presents Kepler, a single interface for designing and executing workflows with minimal control over execution. The thick client provides a web-based portal interface customized based on user tasks. The workflow interface is separate from execution, allowing for execution control capabilities. Provenance tracking is provided, including what actor generated each intermediate result, the path through the workflow for each final result, configuration parameters, association between inputs and outputs, and versions of actors used. Templates and capability-based actors improve workflow generality, and enhanced capabilities such as asynchronous merging and memory-to-memory transfers are supported. Improving usability and extending capabilities for large-scale science workflows are key focus areas for future improvements.
E N D
Our Future with Kepler Terence Critchlow On behalf of the entire SPA team UCRL-PRES-226634
Single interface to design and execute workflow Minimal control over execution Thick client No user customization Independent of workflow Current Kepler User Interface
Separate workflow creation from execution Provide execution control capabilities Web based portal interface Customized based on user task Target Kepler Interface
Provenance Tracking • Provenance director that tracks • Data provenance • What actor generated each intermediate results • What path through the workflow generated each final result • What were the configuration parameters • What is the association between the inputs and outputs • Workflow provenance • Was the workflow based on modifications to a previous w/f • What versions of the actors were used during execution • What workflow engine parameters were used / set
Capability based actors Provide high-level functionality without exposing implementation details Improve generality of workflows Templates Provide overview of specific workflow and let user fill in specific tasks Asynchronous merging Ability to combine outputs from multiple actors into a single stream Memory-to-memory transfers Use multiple ports to support data and control flow between actors Enhanced Capabilites
Conclusions • Kepler is a powerful, flexible workflow infrastructure • In order to directly support scientists we need to focus on improving usability of system • Improve interface and workflow design features • Provide additional capabilities such as provenance • Extend existing capabilities to support large-scale science workflows
This work was performed under the auspices of the U.S. Department of Energy by University of California Lawrence Livermore National Laboratory under contract No. W-7405-ENG-48.