Towards continuous policy-driven demand response in data centers

摘要: Demand response (DR) is a technique for balancing electricity supply and demand by regulating power consumption instead of generation. DR key technology emerging smart electric grids that aim to increase grid efficiency, while incorporating significant amounts clean renewable energy sources. In today's grid, rare event only occurs when actual peak demands exceed the expected peak. contrast, incentivize consumers engage in continuous policy-driven 1) optimize time-of-use pricing 2) deal with variations from non-dispatchable While data centers are well-positioned exploit DR, applications must cope significant, frequent, unpredictable changes available their footprint.The problem challenging since often use distributed storage systems co-locate computation storage, serve as foundation variety stateful applications. As result, existing approaches deactivate servers decreases do not translate well important application-level state may become completely unavailable. this paper, we propose DR-compatible system uses staggered node blinking patterns combined balanced layout popularity-based replication I/O throughput, availability, energy-efficiency varies. Initial simulation results show promise our approach, which increases throughput at least 25% compared an activation approach adjusting real-world wind price fluctuations.