面向应用感知的骨干网缓存方法研究  

Research on the backbone network cache method based on application-awareness

作　　者：韩春静[1,2,3] 杨晔 吕红蕾[3] 葛敬国 李佟[3] 刘韵洁[1] 

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]中国科学院大学,北京100049 [3]中国科学院信息工程研究所,北京100093

出　　处：《通信学报》2017年第12期73-85,共13页Journal on Communications

基　　金：国家重点研发计划基金资助项目(No.2017YFB0801801);国家科技重大专项基金资助项目(No.2017ZX03001019)~~

摘　　要：基于某运营商骨干网上9亿条HTTP请求,分析了HTTP请求组成、内容长度、流行度和时间动态性,提出一种面向应用感知的骨干网缓存方法(AACM,application-aware backbone network cache method)。实验结果表明在缓存空间略有增加的情况下,在线视频的内容命中率提高了15%以上,回源流量和缓存负载这些指标降低了一半左右。Based on over 900 million HTTP requests from an ISP backbone network, the HTTP characteristic of different content types from a wide range of perspectives was analyzed, including request composition, size, popularity and tem- poral dynamics, and an application-aware backbone network cache method was proposed, referred to as AACM. The experiment results show that in the case of a slight increase in cache space, the hit ratio of the online-video increases 15%, and the total source traffic and cache UO load reduced bv about half.

关 键 词：骨干网 HTTP 应用层 缓存 WEB对象 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]