Cache访问优化的空间数据库查询处理技术研究

With the development of computer hardware technology, high-speed chip multi-core processors with shared cache and large memory become the standard configuration of the database server. To take full advantage of these new hardware architectures, it is necessary to improve the efficiency of database query processing on multi-core processors. The database I/O bottlenecks have turned to the emergence of new problems in the memory/cache. Researchers in database fields have proposed some optimization methods to improve memory access performance. But according to spatial database systems, in existing studies, there is little regards for violation problem of multi-threaded parallel execution in the cache access. For these reasons, the computing resources provided by the new hardware architecture cannot be fully used. This project is based on the opportunities of hardware architecture development, and tries to enhance database performance. In this project cache-access optimization for query processing of spatial database will be studied. The cost model will be proposed based on cache performance analysis of the spatial indexing, spatial query operations, which can be servered as theory and experiment basis for efficient query framework design. The spatial parallel query execution framework will be designed based on spatial data access feature. The query processing algorithms will be paralleled, and cache-access optimization algorithms for multi-threaded query execution will be proposed. The research results will be verified by query execution time, cache access missing, parallel execution efficiency, and so on.

随着计算机硬件技术的发展，具有高速Cache的多核处理器和大容量内存成为数据库服务器的标准配置。要充分利用这些新型硬件架构带来的优势，就必须提高数据库查询在多核处理器上的执行效率，解决数据库I/O瓶颈转向内存/Cache后出现的新问题。虽然现有成果已经提出一些改善内存访问性能的优化方法，但针对空间数据库系统，很少考虑多线程并行执行中的Cache访问冲突等问题，无法充分发挥新型硬件结构提供的计算资源。本课题立足于硬件结构发展为数据库性能提升带来的机遇，研究Cache访问优化的空间数据库查询执行优化技术。全面分析空间索引、空间查询等操作的Cache性能，设计代价模型,为高效的查询框架设计提供理论和实验依据。根据空间查询的数据访问特点设计并行查询执行框架，研究空间查询并行优化算法及多线程执行的空间数据Cache访问优化算法。从查询执行时间、Cache访问缺失、并行执行效率等多个方面验证研究成果。

随着计算机硬件技术的发展，具有高速Cache的多核处理器和大容量内存成为数据库服务器的标准配置。要充分利用这些新型硬件架构带来的优势，就必须提高数据库查询在多核处理器上的执行效率，解决数据库I/O瓶颈转向内存/Cache后出现的新问题。虽然现有成果已经提出一些改善内存访问性能的优化方法，但针对空间数据库系统，很少考虑多线程并行执行中的Cache访问冲突等问题，无法充分发挥新型硬件结构提供的计算资源。本项目立足于硬件结构发展为数据库性能提升带来的机遇，研究Cache访问优化的空间数据库查询执行优化技术。项目主要完成的研究工作包括：全面分析空间索引、空间查询等操作的Cache访问性能，设计代价模型，为高效的查询框架设计提供理论和实验依据。根据空间查询的数据访问特点设计并行查询执行框架，研究空间查询并行优化算法及多线程执行的空间数据Cache访问优化算法。从查询执行时间、Cache访问缺失、并行执行效率等多个方面验证研究成果。实现了集群环境下Cache访问优化的内存空间数据库原型应用系统。