Dynamics Analysis Model for Energy Efficient Distributed Storage Systems

摘要

This paper presents a theoretical and expenmental framework based on dynamics to model and analyze energy efficient distributed storage systems. A dynamics analysis model based on Cellular Automata (CA), called DSSCA, is proposed to evaluate distributed Object-based Storage (OBS) systems. DSSCA can have either finite or infinite size with high scalability. Each node represents an Object-based Device (OSD) and can communicate with the neighbor nodes autonomously and perform tasks independently without a centralized control. Moreover, simple energy efficient rules can be adopted by each node and then global behaviors emerge from the evolution and interaction of multiple individuals. To evaluate the intelligence of the system to changes of external behaviors, we proposed a heuristic power-aware mechanism_ Each node has a write request queue, the length of which changes dynamically with access intensity and node throughput In the simulation, we observe an appropriate threshold could both guarantee performance and save power. Under rapid request arrival the entropy of the length of write request queue is near to zero, resulting in no chance of power conservation.