基于DPB^+-Tree的索引复制策略研究

Index Replication Strategy Study Based on DPB^+ -Tree

索引复制是分布并行数据库提供并行性和提高可用性的一个重要手段。本文提出一种适合于索引复制的树结构——DPB^+-Tree,在此基础上研究了相关的索引复制策略,其中副本复制原则考虑了更新/检索比、节点机负载和可靠性需求;索引副本建立允许一个新的副本学习先前的副本;而索引副本更新基于搜索更新机制来完成。对DPB^+-Tree 索引复制策略的仿真实验结果表明,副本对查询的响应性能和负载均衡度有明显改善。

Index replication is an important approach that provides parallel and improves usability of distributed parallel databas. This paper presents a new tree structure - DPB^＋-Tree, which is fit for index replication. Then we research the index replication strategy based on DPB^＋-Tree. The replica duplicating principle includes update/search ratio, machine load and reliability requirement. The replica producing can learn from an old one and the replica updating based on search and update mechanism of DPB^＋-Tree. The simulation results of index replication strategy demonstrate that replica can improve searchresponse characteristic and load balance.