| 基于Cluster结构的多维动态数据分布方法 |
| |
| 引用本文： | 蒋廷耀,睢海燕. 基于Cluster结构的多维动态数据分布方法[J]. 三峡大学学报(自然科学版), 2004, 26(1): 67-72 |
| |
| 作者姓名： | 蒋廷耀 睢海燕 |
| |
| 作者单位： | 1. 三峡大学,电气信息学院,湖北,宜昌,443002
2. 华中科技大学,计算机科学与技术学院,武汉,430074 |
| |
| 基金项目： | 国家863-306资助项目(项目编号为863-306-ZD-01-7) |
| |
| 摘 要： | 数据分布是数据库查询并行处理的基础，良好的数据分布方法对查询性能有着重要影响，本文提出了一种新的基于Cluster结构的多维动态数据分布方法，该方法能保证数据均匀分布在多个处理机上；能动态调整数据片段的大小，使关系始终保持最优并行度；并能有效地支持各属性上的查询操作，性能分析及实验结果表明，在大规模的并行系统中，本文方法的性能优于过去的数据分布方法。
|
| 关 键 词： | Cluster结构 多维数据分布 负载平衡 数据库 |
| 文章编号： | 1007-7081(2004)01-0067-05 |
| 修稿时间： | 2003-04-16 |
A Dynamic and Multi-dimensional Declustering Method Based on Cluster Structure |
| |
| Jiang Tingyao Sui Haiyan. A Dynamic and Multi-dimensional Declustering Method Based on Cluster Structure[J]. Journal of China Three Gorges University(Natural Sciences), 2004, 26(1): 67-72 |
| |
| Authors: | Jiang Tingyao Sui Haiyan |
| |
| Abstract: | Data declustering is the foundation of the parallel processing of database query, which has significant effect on query performance. In this paper a new dynamic and multi-dimensional declustering method based on cluster structure called as BMAP has been presented;this method can keep data distributed to multiple processors evenly and dynamically adjust data block to partition relation properly that guarantee relations always keep the optimal parallelism and support different query operations on all attributes. The new method records the data distribution on assistant attribute to improve the query efficiency. The results of performance analysis and experiment evaluation show BMAP has better performance than traditional data declustering methods when the scale of parallel system is large. |
| |
| Keywords: | cluster structure parallel processing multi-dimensional data declustering load balance |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
