MFCuts多维报文分类算法  

MFCuts: Multi-field packet classification algorithm

作　　者：李育强[1] 姚凯[1] 李常春[2] 兰云海 

机构地区：[1]电子科技大学计算机科学与工程学院,成都610054 [2]重庆通信学院信息工程系,重庆400035 [3]中南空管局通信网络中心,广州510470

出　　处：《计算机应用》2014年第A01期169-172,共4页journal of Computer Applications

基　　金：国家发改委2012年下一代互联网技术研发;产业化和规模商用项目([2012]1763);教育部-中国移动科研基金资助项目(MCM20123041;MCM20136061);重庆市应用开发计划项目(CSTC2013yykfA40003)

摘　　要：在超高速网络中的流量分析和统计面临极大的性能挑战,而流分类算法是影响性能的关键因素。为改进Hicuts和HypterCuts的不足,提出了一种新的多维分类算法MFCuts。其根据一个给定的算法的优劣度量公式,找出所有唯独不同分割中最优的一个进行实际分割。实验结果表明,它能克服同类算法的不足,有效避免了HiCuts每次只能选取一维进行分割以及HypterCuts分割具有较大随意性的特点,有效提高流分类速度。The analysis and statistics of flow in ultra-high-speed network face great challenges in performance to which the algorithm of packet classification is the key point. A new algorithm named MFCuts based on the analysis of some well- known algorithms was proposed to overcome the defects of Hicuts and HypterCuts. MFcuts find the optimal partition in aU splits of different dimensions based on the merits of a given mensurement formulas. According to optimal one, MFCuts makes the actual split. The results show that the MFCuts could overcome the defects of other similar algorithms, effectively avoid the disadvantage of Hicuts which can only select one-dimensional cut and HyterCuts cuts with large randomness, and could improve the classification speed efficiently.

关 键 词：报文分类 分类算法 流量分类 多维分类 决策树 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]