论文部分内容阅读
在传统数据库系统中,I/O开销一直是连接算法性能瓶颈.固态盘具有随机读写性能高和低延迟的特性,使得利用固态盘改善连接算法I/O性能成为可能.提出基于硬盘和固态盘混合存储系统下的组合盘GRACE散列连接算法(CGHJ).CGHJ对传统GRACE散列连接算法分区阶段作如下改进:将固态盘作为内存和硬盘之间的缓存,位于内存的各个散列桶先写到位于固态盘的临时文件缓存;临时文件缓存满后,利用迁移缓存将固态盘中的内容转移到位于硬盘的各个分区文件.新的分区方式增大了分区过程中随机写到每个分区文件的块,减少硬盘分区过程中产生的随机I/O.实验结果显示在缓存容量较小或表的数据规模较大情况下,CGHJ能取得很好消除硬盘随机I/O的效果.
In the traditional database system, the I / O overhead has always been a bottleneck in the performance of the connection algorithm.Solid disk has the characteristics of high random read and write performance and low latency, which makes it possible to improve the I / O performance of the connection algorithm by using the SSD. GRACE hash join algorithm (CGHJ), a composite disk under a hybrid storage system CGHJ makes the following improvements to the traditional GRACE hash join algorithm partitioning stage: The SSD is used as a cache between memory and hard disk, located in each hash bucket of memory The first write to the temporary file cache in the solid state disk; temporary file cache is full, the use of migration cache SSD content is transferred to each partition in the hard disk file.New partitioning method increases the partitioning process to write each Partition file block to reduce the random I / O generated during hard disk partitioning.Experimental results show that CGHJ can effectively eliminate hard disk random I / O when the buffer capacity is small or the size of the table data is large.