一种可扩展机群文件系统的设计与实现

介绍了可扩展机群文件系统TH-CluFS的设计与实现．TH-CluFS基于共享文件模型，采用无服务器的对等体系结构．同时，为了有效利用杌群系统累积的资源，TH-CluFS采用了复制与分布相结合的元数据分布管理策略、文件粒度的数据分布存储机制，以及一种茚治的I／O负载平衡方法．实验结果表明，TH-C1uFS很好地统一了机群文件系统在单一系统映像、高性能以及可扩展性等方面的需求．

Design and Implementation of a Cluster File System

The paper presents the design and implementation of TH CluFS cluster file system. TH CluFS employs a serverless architecture based on file sharing model. What is more, to make full use of the accumulated resources in the cluster, we introduce a distributed metadata management policy which cooperates replication and fragmentation effectively, a file distribution mechanics at file granularity and an autonomous I/O load balancing method into TH CluFS. According to the experiment result, we conclude that TH CluFS fulfils the requirements of single system image, performance and scalability gracefully.