基于主动探测的不确定环境下IP网丢包率推理机制研究

Packet loss rate is an important indicator of network performance. How to monitor the link loss rates in real time, and detect and locate the lossy links in time is one of the key technologies to ensure network reliability. However, as the network scale and heterogeneity increasing, the environments of IP networks are becoming more uncertain. Hence, conventional link loss monitoring methods cannot ever applicable to today's network environments. In this project, we focus the research on probe based link loss inference mechanism in uncertainty IP networks. First, an adaptive probe selection algorithm is investigated to select the least and reliable probe path whose loss rates will be measured. Secondly, an efficient loss inference algorithm based on end-to-end loss rates is explored to deal with the uncertainty factors. Finally, the proposed mechanism will be verified by large scale IP network simulator. We also apply the new mechanism to the real Internet supported by Planetlab to prove its practicability. The new proposal in this project is expected to provide theoretical references for improving the effect of monitoring and managing today’s and the future IP networks.

丢包率是衡量网络性能的重要指标，对全网链路丢包率的实时监测，及时发现和定位丢包异常的链路是保证网络可靠性的关键技术之一。而随着IP网规模的扩大，异构性的增强，网络环境变得越来越不确定，现有的基于确定网络环境的链路丢包率监测技术已越来越不能适应网络的快速发展变化。因此，本项目针对不确定IP网环境下的链路丢包率推理机制展开研究，拟提出一种基于主动探测的链路丢包率推理方法。首先，研究一种自适应的探测路径选择算法，选择尽量少的可靠路径并测量端到端路径丢包率；其次，研究一种基于端到端丢包率的链路丢包率推理算法，在能够处理网络中的不确定因素的同时，保证计算准确度和计算效率。最后通过大规模IP网络仿真软件对新算法进行正确性验证，并在Planetlab提供的真实Internet站点中对算法进行可行性验证。为更有效地监测和管理现有的和未来IP网性能提供理论支撑。

本课题的研究目标是基于主动探测的不确定IP网络环境下链路丢包率推理机制。课题组成员围绕着该目标，经过三年的研究工作，取得了一系列的研究成果。代表性的研究内容包括：1）提出了一种自适应的端到端探测路径选择算法；2）提出了故障多发环境下的自适应路径选择算法及IP网链路丢包率推理算法；3）提出了不稳定IP网络环境下的链路丢包范围推理算法；4）提出了不确定IP网络环境下的链路丢包率推理算法；5）设计并开发了基于真实ISP拓扑的大规模不确定IP网络仿真平台。. 在本课题的资助下，项目组成员取得了以下研究成果：1）在国内外知名期刊上发表了学术论文4篇，其中SCI检索论文2篇；2）在领域内顶级国际会议INFOCOM（将于2018年4月在美国召开）和知名国际会议ISCC上共发表学术论文2篇；3）提交国家发明专利3篇；4）培养硕士研究生2名。