Apply Aggregate I/O to Improve Performance of Network Storage Based on IP

Abstract

Network storage provide high scalability, availability and flexible for storage system, and are widely applied to many fields. Small I/O heavily effects performance of storage system include of network storage. Current several methods to solute this problem need theory model and but only improve write performance. In this paper, the aggregate I/O model is presented first. Moreover, we also illustrate how to apply this methodology in NAS system based on IP network. The aggregate I/O module called CFS (Chunk File System) is designed in both NAS device and customer end. In our experiment, the analysis of results shows correctness of model and an improvement made by aggregate I/O for network storage performance. Our experimental results show that CFS using aggregate I/O can improve throughput by factor of 2-4 in same response time without it.

This paper is supported by both National Natural Science Foundation of P R China under the Grant NO 60173043, NO 60273073 and National 973 Great Research Project of P R China under the Grant NO G1999033006.