一种改进的混合范围划分方法

An Improved Hybrid Range Partitioning Strategy

混合范围划分方法[1]给出了计算数据分置节点数的公式以及数据划分的方法;加强的混合范围划分方法[2]通过引入可变范围的数据分块,达到了节点间数据存储量的一致,解决了混合范围划分方法的数据倾斜问题。为了达到系统运行时节点间的查询负载平衡,本文对上述方法进行了改进,引入了热度的概念来反映查询负载,通过对系统初始阶段的数据划分方法和系统运行阶段的数据迁移方法的改进,来达到上述目的。

Hybrid range partitioning strategy introduces a formula to compute the amount of nodes to distribute data and the data partitioning strategy based on identical range sizes then, an enhanced hybrid range partitioning strategy achieves data storage balancing and solves data skew between the nodes in the parallel real-tlme database system by varying range sizes. But, in order to achieve the query load balancing during the system runtime, the strategies above should be improved. An improved hybrid range partitioning strategy introduces the definition of Heat on behalf of query load. What＇s more, it introduces the data partitioning strategy during system initialization, and data migration strategy during system runtime.