Adaptive power management with fine-grained delay constraints

Power consumption of computing systems has become an important issue in industrial design areas. Dynamic power management (DPM) is an effective way to low the system power. This paper presents a new expert-based heuristic algorithm with fine-grained delay constraints (FDC-DPM), which selects the best policy from a set of well-known policies dynamically. Three rules are presented for FDC-DPM to make a choice. FDC-DPM gives a flexible way to make a good tradeoff between energy consumption and fine-grained delay constraint. Compared with the machine learning method in [2], FDC-DPM is simpler and it can achieve comparable energy savings with the same delay constraints under different workloads.