基于层次结构的远程数据复制系统

针对传统的集中式复制结构存在的单点瓶颈和扩展性差等不足，提出了基于层次化结构的数据复制系统。通过将复制节点之间的拓扑结构抽象为图，提出了使用最小代价树的方式对节点进行组织。通过将数据由树根逐层扩散复制到树叶，使得在保证数据一致性的同时，分散了主节点的负荷。最后分析了使用复制数据进行灾难恢复的过程，并通过原型实验对集中式与层次化复制结构进行了比较。

Hierarchical Structured Remote Data Replicating System

The hierarchical structure of data replication, which organizes all replicated nodes into a tree, is adopted to overcome the side effects introduced by the traditional type of centralized replication, such as critical elements and weak scalability. This paper describes the organization of the nodes as the minimum spanning tree since the topology of the system can be treated as a graph. By propagating data from the root to all the leaves, the system distributes the load of the original node into the whole system as well as guarantees the data consistency all nodes. In addition, the method of data recovery is discussed. Finally a comparison between centralized system and hierarchical one is illustrated.