基于二维滑动窗口的数据网格通信模型

在由多计算机集群构成的数据密集型网格环境下,为了解决通信峰值阶段的通信冲突问题,提出了一个基于二维滑动窗口机制的通信模型。给出了由多计算机机群构成的数据密集型网格数据节点DGDN(data-intensivegriddatanode)、完全图通信和二维滑动窗口等一系列形式化定义;利用计算节点和网络的通信能力,研究了二维滑动窗口机制,描述了单个机群内的完全图通信算法CA2DSW(communicationalgorithmbased2D-sliderwindow)以及双机群内的计算机之间的完全图通信算法CADC(communicationalgorithmindoublecluster)。基于网络距离机制、CA2DSW和CADC,设计了DGDN上的通信模型CMDGDN。理论分析和实验结果表明,该模型有效地解决了数据网格环境下通信峰值所造成的并行计算效率低下的问题。该模型适合于基于网格的数据密集型并行计算。

Data grid communication model based on 2D-slider windows

For resolving the conflict problem of the peak communication in data-intensive grid that composed of com-puter clusters, a communication model based on 2D-slider window was discussed. A series of formal definitions, such as DGDN (data-intensive grid data node), the complete graph communication and 2D-slider window(2DSW), were pro-posed. By using the abilities of computing nodes and networks, the 2DSW mechanism was studied. The complete graph communication algorithms based on 2DSW (CA2DSW) in the single computer cluster and the complete graph commu-nication algorithm in double clusters (CADC) were described. Basing on network distance, CA2DSW and CADC, the complete graph communication model in DGDN (CMDGDN) was designed. The analysis and experiment results show that CMDGDN effectively resolve the low efficiency during the peak communication in grid. It can be fit for the data-intensive parallel computing based in grid.