高性能互连网络故障、管理与测试探讨  

Discussion of the Fault, Management and Testing Issues for High Performance Interconnection Networks

作　　者：曹继军[1] 徐金波[1] 徐佳庆[1] 

机构地区：[1]国防科学技术大学计算机学院,长沙410073

出　　处：《高性能计算技术》2014年第1期35-41,共7页

摘　　要：高速互连网络是高性能计算系统的关键部件，其易管理性直接影响整个系统的RAS（可靠性、可用性和服务性）特性。本文首先研究了网络故障——对网络故障的分类进行了探讨，重点研究了无丢弃网络数据链路层故障的类别和原因，提出了降低网络故障影响的技术思路；然后讨论了网络管理——总结了网络管理的目标和需求，探讨了网络管理的实现，提出了3种提高网络管理易用性的实现方法；最后研究了网络测试——将现有的网络测试方法分类为白盒、黑盒和灰盒测试方法，分析比较了各种测试方法的优缺点。本文探讨的相关结论对于设计易管理的高速互连网络具有重要的参考价值。High-speed interconnect network plays an important role in high performance computing system And its manageability directly affects the RAS （Reliability, Availability and Serviceability） of the whole system. This paper first analyzes the network fault. The classification of network fault is discussed, where the types and causes of fault in data-link layer of lossless network is emphasized. Some technical ideas are proposed to reduce the effect of network fault. This paper further discusses the network management, including summarizing the goals and requirements of network management and discussing its implementation, and then proposes three technical ideas to improve the usability of network management. Finally, the network testing is studied. This paper divides the approaches of network testing into three categories, i.e. white-box testing, black-box testing and grey-box testing. The advantages and disadvantages of the three network testing approaches are compared respectively. The ideas that are discussed and proposed in this paper will guide the design of manageable high speed network in the future.

关 键 词：高速互连网络 网络故障 网络管理 网络测试 故障定位 故障诊断 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]