区域分片下的分布式空间查询处理与并行调度方法

Distributed Spatial Query Processing and Parallel Schedule Based on Zonal Fragmentation

针对区域分片的数据分布特征,在已有的跨边界连接转化及优化规则的基础上,提出了一种先静态任务分配、后动态任务调整的混合并行调度策略。实验表明,该混合策略有效解决了分布式空间并行查询的负载平衡问题,进一步提高了分布式空间查询的效率。

According to the characteristics of zonal fragmentation and the existing methods oftransformation and optimization based on cross-border topological join, we propose a hybrid parallel scheduling strategy with the principle of static and dynamic combination. Firstly, the excutive sites of the fragment ioins are determined using the Move Small rule. Seconde- ly, the pre-allocated task queue on each site is reranked on the principle of larger area of FM- BR priority. Finally, the fragment join tasks in each queue are redistributed in order to im- prove load balance by comparing the costs of three typical task sheduling cases. The experi- mental result shows that this method solves the load-balancing problem in effectively and further improves the performance of distributed spatial query.