Data Placement Strategy in Data Center Distributed Storage Systems

The purpose of research data placement strategy is to improve the fault tolerance of data center systems. However, commonly used strategies are not considering the transmission time to repair the failure nodes and load balancing together. In this paper, we propose a data placement strategy applied to BCube data center network topology, which has jointly taken data access efficiency, the recovery time and the load balance into consideration. According to the strategy, the system selects storage nodes based on the evaluation. By comparing with the HDFS placement strategy, the data placement strategy we proposed not only achieves load balance, but also ensures repairing failure nodes rapidly.