Hybrid Storage System Power Optimization

Abstract

This paper investigates an energy-efficient hybrid storage system framework. An integer linear programming (ILP) approach is used to formulate power optimization frameworks for each individual storage system by separately developing appropriate system constants, binary decision variables, and associated constraints for a four power level Mobile Random Access Memory (MRAM) and a three power level Phase Change Memory (PCM). Subsequently, a hybrid configuration of MRAM and PCM storage systems is used to describe corresponding system variables and associated constraints under predefined power-level operation conditions to quantify actual power usage for programs with plausible execution patterns. The simulation result illustrates the efficiency of the proposed hybrid storage power optimization ILP framework.