一个并行查询优化器的设计与实现

Design and Realization of a Parallel Query Optimizer

并行查询优化器的目标是缩减庞大的计划搜索空间,获得优化的查询规划。为此,并行实时数据库PRTD-BASE查询优化器针对无共享结构(SN),充分考虑通信开销,采用两阶段优化方法,依据代价估计模型先对查询树进行基于代价估计的顺序优化,然后利用启发式规则对顺序优化的查询计划进行并行化,充分利用了多处理机的并行性,获得了较快的查询响应时间。

The key problem is how to prune the big search space in order to get an optimized query plan for query parallel optimization.Therefore,based on the shared-nothing structure and taking communications into account with cost estimation,a two-phase optimization strategy is used to optimize the query tree sequentially in the first phase and to parallelize the sequential optimization tree in the second phase by heuristic rules in the PRTDBASE optimizer.Multi-processor parallelism is utilized and faster response time is obtained.