Effective Dynamic Voltage Scaling Through CPU-Boundedness Detection

Dynamic voltage scaling (DVS) allows a program to execute at a non-peak CPU frequency in order to reduce CPU power, and hence, energy consumption; however, it is oftentimes done at the expense of performance degradation. For a program whose execution time is bounded by peripherals' performance rather than the CPU speed, applying DVS to the program will result in negligible performance penalty. Unfortunately, existing DVS-based power-management algorithms are conservative in the sense that they overly exaggerate the impact that the CPU speed has on the execution time. We propose a new DVS algorithm that detects the CPU-boundedness of a program on the fly (via a regression method on the past MIPS rate) and then adjusts the CPU frequency accordingly. To illustrate its effectiveness, we compare our algorithm with other DVS algorithms on real systems via physical measurements.