Power-aware Online Dynamic Scheduling for Multiple Feasible Intervals

Multiple Feasible Intervals (MFI) is a task model which has more than one interval for each task instance to run. In our previous work, different scheduling algorithms are proposed for both static MFI dynamic MFI during run-time. Real-life situations demand dynamic MFI, but calculating an MFI schedule is a NP-Hard problem and efficiency of algorithm itself is one of the main concerns. In this paper, we introduce a more efficient online dynamic algorithm, Online Dynamic Multiple Feasible Intervals (ODMFI) which aggressively reduces power consumptions with help from accurate performance prediction models and integration of a simple caching policy into scheduling algorithm to cope with changing feasible intervals during run-time. Our simulations show that ODMFI generates schedules which have close level of energy consumption to Power-Aware MFI [3] with less overhead compared with Dynamic MFI [6]. * This work was supported in part by the National Science Foundation under Award No. 0720856.