While a vast number of tools are available for service providing managers (SPMs) to schedule jobs on grids, there is a lack of generally applicable methods for reducing energy consumption while ensuring good quality of service. In this project, one of our primary goals is to design an arsenal of innovative algorithms for energy-aware scheduling and mapping of workflows in a service-oriented grid environment. Typically, grid workflows are represented as task graphs, containing ample idle or slack periods for which the CPU can be throttled using the DVS (Dynamic Voltage Scaling) technique. Our DVS-based scheduling algorithms will be built around a theoretical framework called DYNAMO (DYNamic self-Adjusting Multi-objective Optimization) for simultaneous optimization of energy and performance. The arsenal of novel static and dynamic scheduling algorithms in the DYNAMO framework provides SPMs with maximum flexibility in trading performance for energy to suit a given application scenario.
For self-adjusting control, our other primary goal is to develop the GridPac middleware environment that will leverage the full arsenal of algorithms to schedule parallel applications across a distributed grid for system-wide optimization. The middleware will allow SPMs to specify policies that will enable the scheduling of jobs onto the computational resources in a self-optimizing manner for achieving energy minimization and performance maximization. The middleware will include effective energy monitoring techniques, task dispatching and control messages, as well as the provision for policy adjustment.