一种层次化网络距离预测机制

A Hierarchical Network Distance Prediction Mechanism

网络坐标系统向分布式应用提供了一种高效的网络距离信息获取机制,但现有基于单一度量空间嵌入的距离预测机制难以精确描述因特网复杂的层次结构特征,进而导致较大的距离预测误差.文章提出了一种分域的层次化网络距离预测机制NetPharos,该机制根据因特网结构以及性能特征将其划分为相互独立的核心预测域和边缘预测域,通过相关边缘预测域和核心预测域内预测值迭加获取节点间距离信息.理论分析和仿真实验显示,NetPharos能够有效解决预测过程中短距离和长距离的相互干扰问题,提高预测精度.

Network Coordinate system provides an efficient mechanism to obtain network distance （latency） information with limited times of measurement. However, prediction mechanisms based on single metric space embedding cannot describe the complex hierarchical structure of In- ternet precisely, and result in great prediction errors. By analyzing the hierarchical structure fea- ture of Internet, a region-partition based hierarchical network distance prediction mechanism named NetPharos is proposed, which divides Internet into independent core prediction region and many edge prediction regions based on the structure and performance feature of Internet. Dis- tances between any two nodes are obtained by accumulating distances in their edge regions and the core region. Theoretic analysis and simulation results show that NetPharos can avoid the interference between short distances and long distances during distance prediction process, and improve distance prediction accuracy.