摘要
为了提高系统的可扩展性及整体查询速度,云存储中通常将元数据与数据进行分开管理。在小文件存储系统中,集群规模的增大会导致存储元数据的主节点性能瓶颈。该文针对此问题,在小文件合成大文件技术的基础上,应用基于Vandermonde矩阵的RS算法改进了其冗余机制,减少了系统存储的元数据。引入group概念,对主节点应用元数据形成集群进行管理,并借鉴树型数据结构,设计了基于group的树型元数据结构,同时采用主、从节点共同存储元数据的方式,分散单一主节点存储元数据的压力。
To improve system scalability and the overall query speed, cloud storage is typically metadata and data are managed separately. In a small file storage system, increasing the cluster size can cause performance bottlenecks master node storing metadata. Aiming at this problem, on the basis of small files into a large file technology, application- based RS algorithm Vandermonde matrix to improve its redundancy mechanisms to reduce the metadata stored in the system. The introduction of group concept, the main application metadata form a cluster node management, and draw a tree data structure, a group of tree - based metadata structures, while using the main way from the common node storing metadata, decentralized single master node pressure storage of metadata. Conclusions show that the proposed scheme effectively improves the cloud storage from the main structure master and overall system performance.
出处
《电脑知识与技术》
2016年第6期118-121,共4页
Computer Knowledge and Technology
关键词
云存储
元数据结构
元数据存储策略
树型数据结构
cloud storage
metadata structure
metadata storage strategy
tree data structur