1（Department of Computer Science and Technology，Hangzhou Dianzi University，Hangzhou 310018，China）2（The Network Media Cloud Processing and Analysis of Zhejiang Engineering Technology Research Center，Hangzhou 310018，China）3（The Complex System Modeling and Simulation of the Key Laboratory of Ministry of Education，Ministry of Education，Hangzhou 310018,China）4（College of Electrical Engineering，Zhejiang University，Hangzhou 310058，China）
Abstract：With the increase in the scale of block-level cloud storage services caused by cloud computing system，the load of cloud storage subsystem also increases，which seriously affects its I／O performance.To address above problem，this paper designs and implements a block-level network disk cache system，named MN-cache，which fits for multi-node condition.Firstly，this paper designs the system architecture of MN-cache system，including the metadata server to process metadata and the cache client to implement cache function.Secondly，base on the Linux Device Mapper mechanism，this paper implements the client cache mechanism，in which regards the local cache disks of cache client nodes in the LAN as shared cache and improve the performance of cloud storage subsystem through the advantage of the LAN and local cache disk.Finally，according to the time-consuming of each cache stages，this paper analyzes the performance of MN-cache system，and validates the analysis through experiment.The experimental results show that：in case of poor I／O network storage system，MN-cache system can improve the overall performance of storage system.
2016, Vol. 37 
