针对互联网Web应用特别是Web2.0应用的特殊存储需求,提出一种基于容器的自组织存储模型(CSS-M).CSS-M使用容器管理存储空间,聚簇存放用户文件,提高用户数据备份、迁移和恢复效率.一方面模型通过文件唯一标识提供文件的快速访问,另一方面它采用文件集来组织用户文件成树状,提供灵活的文件管理功能.模型使用容器作为数据定位和复制的基本单元,并利用对等覆盖网络技术自组织地维护容器元数据.另外,主从容器复制技术和基于容器状态的容器恢复技术保证了数据的可靠性和一致性.利用CSS-M实现了一个存储系统原型,初步的实验结果表明,CSS-M提供良好的性能和可扩展性,能够满足互联网WEB应用的存储需求.
A container-based self-organizing storage model (CSS-M) was presented for web2.0 applications. Based on the concept of container,CSS-M clustered the user file together on disk to explore the file access locality and enable the fast data backup,migration and recovery. CSS-M employed ID based access interface to boost the access efficiency,and provided the flexible file management by using collections to organize files into trees. CSS-M maintained the container metadata as soft-state and distributed the metadata across all storage nodes using peer to peer technology. Data reliability and consistency were guaranteed by using the primary-copy replication and the state-based container recovery. Experimental results show that CSS-M provides high scalability and throughput in file operation services.