Load Balance Analysis-Based Retrieval Strategy in a Heterogeneous File System

In order to adapt to heterogeneous storage environment and improve the overall I/O efficiency and system throughput, we address the optimization of replica management and retrieval strategies in a distributed file system. The proposed strategies can enhance the system's adaptability to complex environment, improve the system's rationality for load balance, and reduce the overall storage cost of the system. Specifically, this paper presents a performance metric of measuring the load of nodes. Firstly, we address the concept of comprehensive load, and then propose the method of computing comprehensive load, which is based on the multi-dimension analysis on the system. In addition, we propose the replica management and retrieval strategies, which consider the comprehensive load of nodes in the distributed file system, and optimize the allocation of loads among nodes systematically. Based on the abovementioned strategies, we address the replica placement strategy, the replica management strategy, and the retrieval algorithm in a distributed file system, in consideration of the heterogeneity of nodes in the cluster, the difference between files, and the real-time performance of nodes. All these strategies and algorithms can provide an optimization of replica and retrieval process in a distributed file system.