多媒体通信业务流识别与分类方法综述  

Survey of Methods for Traffic Identification and Classification in Multimedia Communications

作　　者：董育宁[1] 王再见[1,2] 房曙光[1] 张健[1] 

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003 [2]安徽师范大学物理与电子信息学院,安徽芜湖241000

出　　处：《南京邮电大学学报（自然科学版）》2013年第3期35-44,共10页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(60972038;61271233);教育部高等学校博士学科点专项科研基金(20103223110001);2011年度江苏省研究生培养创新工程(CXZZ11_0396);安徽高校省级自然科学基金(KJ20112140)资助项目

摘　　要：准确、高效的业务流识别与分类是保障多媒体通信端到端QoS(Quality of Service)、执行相关网络操作的前提。但多媒体通信业务构成复杂、具有较严格的QoS约束,且在包/流水平统计特征多样性,业务统计特征有效选取直接关系到识别和分类方法的有效性。在介绍相关研究成果的基础上,文中从业务特征角度对现有技术进行分类,进而对比各类方法的性能,同时在探讨当前业务流识别方法存在对新业务识别准确度不高、实时性不足等问题的基础上,结合跨域QoS类映射弹性需求的特点,给出跨域QoS类映射中多媒体业务识别架构。整个架构的目标是准确、高效地识别多媒体流,为聚集流的形成做好前期准备,为保障高效的端到端QoS提供技术支撑。最后,总结了发展趋势和面临的挑战。Accurate identification and categorization of multimedia traffic is the premise of end-to-end QoS（Quality of Service）guarantees.This paper introduces related works on identification and categorization of multimedia traffic,classifies current identification methods from the point of view of traffic characteristics,and contrasts the performances of some typical identification methods.Meanwhile,by studying the shortcomings of current identification methods on identifying new services and analyzing the requirement of QoS class mapping for different QoS domains,this paper presents a new multimedia traffic identification framework for QoS class mapping across different QoS domains.The objective of the proposed framework is to ensure accurate and efficient identification and categorization of multimedia traffic,to prepare the aggregation of multimedia flows,and to lay a technological foundation of efficient end-to-end QoS support.Finally,some possible future trends and challenges are given.

关 键 词：多媒体通信 流识别 流分类 QOS 机器学习 

分 类 号：TN919.8[电子电信—通信与信息系统]