Back Propagation Neural Network Based Data Adaptive Storage Method for Heterogeneous Storage Systems

Heterogeneous storage devices are commonly used in nowadays storage systems. In this paper, we propose a new Back Propagation neural network based Data adaptively storing Method (BPDM), in order to adapt the heterogeneous devices by storing the frequently accessed data in immediate future in fast devices as much as possible. The experiment results show that BPDM performs well on heterogeneous devices based storage systems. In statistic, BPDM gains up to 33.2% and 49.8% less average responding time and average serving time compared to other existing methods.