基于节点动态内容流行度的缓存管理策略  

A Strategy Based on Dynamical Content Popularity for Cache Management

作　　者：张果[1] 汪斌强[1] 张震[1] 梁超毅 

机构地区：[1]国家数字交换系统工程技术研究中心,河南郑州450002 [2]解放军信息工程大学,河南郑州450001

出　　处：《电子学报》2016年第11期2704-2712,共9页Acta Electronica Sinica

基　　金：国家自然科学基金创新研究群体项目(No.61521003);国家973重点基础研究发展计划(No.2012CB315901;No.2013CB329104);国家自然科学基金(No.61372121;No.61309019;No.61309020;No.61572519);国家863高技术研究发展计划(No.2015AA016102;No.2013AA013505)

摘　　要：针对命名数据网络中节点无法感知内容流行度变化的缺陷,提出了基于缓存内容流行度动态变化的内容管理策略.将缓存分为主缓存(Primary Cache,PC)和副缓存(Secondary Cache,SC),分别用于识别和保护流行内容;采用标准布鲁姆过滤器(Standard Bloom Filter,SBF)过滤流行内容请求;引入滑动时间窗口算法和HASH表对副缓存内容进行细粒度的统计分析,进而管理缓存内容.仿真显示,与现有算法相比,该策略以增加少量复杂度为代价,延长高流行度内容的缓存驻留时间,提高了缓存命中率,减轻了服务器负载,并具有可扩展性,具备单线路40Gbit/s的报文处理能力.To overcome the drawback that nodes in Named Data Networking are insensitive to the change of the content popularity, a dynamic content popularity based cache management strategy is proposed. The strategy divides the cache into pri- mary and secondary one. The former is used to identify popular content and the latter is used to protect it. Standard Bloom Fil- ter is adopted by the strategy to filter popular content requests. The strategy also introduces sliding window and hash table to analyze the content of secondary cache in fine granularity and manage the cache content. Simulation results show that, com- pared with traditional strategies, our algorithm prolongs the cache residence time of high popularity content, increases cache hit ratio and reduces server loads. Our algorithm is also scalable and has the ability to process packets at 40Gbit/s.

关 键 词：命名数据网络 动态内容流行度 线速 内容管理 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]