基于骨干网上行流量深度报文检测的网络质差评估方法  

Evaluation of network quality difference based on deep message detection of backbone network traffic

作　　者：余雪芬 Yu Xuefen(China Telecom Co.,Ltd.,Fuzhou Telecom Branch,Fuzhou 350001,China)

机构地区：[1]中国电信股份有限公司福州分公司,福建福州350001

出　　处：《无线互联科技》2020年第22期37-39,共3页Wireless Internet Technology

摘　　要：文章介绍通过基于骨干网上行流量深度报文检测(Deep Packet Inspect,DPI)的网速慢评估方法。DPI包含了报文信息的七层解析,也就包含了TCP协议的四层交互信息。TCP连接建立异常、TCP交互重传或者乱序的统计量都能表征网络质量情况。但对于运营商骨干网来说,上下行流量特别大,特别是下行流量,因此,运营商也只能存储上行若干TCP端口流量。由于缺乏双向TCP流量特征,网络质量异常很难精准发现。文章通过研究骨干网上行流量HTTP/HTTPS端口与用户申告网速慢的关联关系,提出骨干网上行流量特征的生成方法及网络质差评估方法。This paper introduces the network speed slow evaluation method based on backbone network traffic depth message detection(Deep Packet Inspect,DPI).DPI contains seven layers of parsing of message information,which includes four layers of interactive information of TCP protocol.Statistics TCP establishing anomalies,TCP interactive retransmissions,or disorder can characterize network quality.However,for the backbone network of operators,up and down traffic is particularly large,especially downlink traffic,so operators can only store a number of uplink TCP port traffic.Due to the lack of bidirectional TCP traffic characteristics,it is very difficult to accurately find abnormal network quality.By studying the relationship between the traffic HTTP/HTTPS port of backbone network and the slow speed of user s declaration network,this paper puts forward the generation method of traffic characteristics and the evaluation method of network quality difference.

关 键 词：骨干网 深度报文检测 网络质差评估 

分 类 号：TN915.08[电子电信—通信与信息系统]