基于DPB+-Tree的数据迁移策略研究
Data Migrating Strategy Study Based on DPB+-Tree
-
摘要: 提出了一种适合于数据迁移、又能保证分布并行特性的树结构DPB+-Tree,讨论了基于DPB+-Tree的数据迁移策略,其中数据节点迁移采用分布式提交协议来保证原子性,索引重构通过对溢出链的hash重排来实现,迁移算法则通过设置负载系数的两个阈值来对负载倾斜进行判断。经模拟实验结果表明,该数据迁移策略能够有效改善系统的负载均衡和吞吐率特性。Abstract: This paper presented a new tree structure DPB+-Tree, which suitable for data migrating and distributed and parallel. We have studied the data migrating strategy based on DPB+-Tree in which include data node migrating, index restructuring and migrating algorithm. The data node migrating utilise distributed commit protocol to ensure the atom-operation characteristic. Index restructing rearrange overflow chain list following hash rule. According to set two thresholds, a recursive migrating processing is triggered when load coefficient exceed limit and there is another machine can receive data. The simulation results demonstrate:data migrating strategy can improve load balance and throughput characteristic.