摘要： 在大型多维数据集合处理中，对多维数据集合的拆分及其在磁盘上的存储分配是重要的研究课题。由于磁盘的机械运动已形成了数据I/O时的速度瓶颈，因此通过采用并行I/O技术，将多维数据进行有效的拆分，并在多个磁盘间进行分布存储是克服瓶颈的有效办法。基于此，论文中提出了一种多维数据的循环拆分方法，它是对二维数据集合循环拆分分配方法的扩展，性能比较与分析表明了新算法的有 效性。
关键词:
多维数据集,
并行I/O,
数据拆分,
循环分配策略
Abstract: It is an important research topic to process the large-scale multidimensional dataset by program. When the data is input or output, the speed of reading and writing is limited by the mechanism movement of disks itself. The parallel I/O technology is efficient to eliminate the I/O bottleneck. Parallel I/O can decluster the multidimensional dataset and store them. Based on this fact, a new cyclic declustering strategy for the large-scale multidimensional dataset is proposed. The experiment shows its validity.
Key words:
Multidimensional dataset,
Parallel I/O,
Data declustering,
Cyclic distributed strategy
中图分类号:
曾碧卿;陈志刚. 并行I/O中大型多维数据集合分配策略研究[J]. 计算机工程, 2006, 32(19): 38-39,4.
ZENG Biqing; CHEN Zhigang. Research on Declustering Strategy of Large-scale Multidimensional Dataset of Parallel I/O[J]. Computer Engineering, 2006, 32(19): 38-39,4.