一种基于区域划分和内容流行度的NDN缓存策略

Cache strategy of named-data networking based on community nodes division and content popularity

针对命名网络中路由器缓存能力有限的问题,提出了一种基于区域划分和内容流行度的NDN缓存策略(CCNCP).该策略根据拓扑结构划分区域,并综合衡量节点重要性选出局部中心节点和全局中心节点;引入分类的内容流行度,合理地将不同流行度的内容缓存在不同重要度的节点.首先采用经典的GN算法进行社区划分.定义了度中心性、紧密中心性、介数中心性等中心性度量指标.然后建立了系统模型,给出了缓存策略.最后对Zachary空手道俱乐部网络的社团结构进行改编,并引入真实拓扑GARR,使用Icarus模拟器进行仿真试验.结果表明:CCNCP在2个不同网络结构中表现得最好;当参数S等于0.25时,CCNCP和第2好缓存策略ProbCache的缓存命中率、延迟、路径拉伸比分别为30.9%和26.8%、52.9和56.2 ms、0.584和0.617,前者比后者分别提高4.1%、降低5.9%、降低5.3%,可见CCNCP在3个指标上与现有的典型策略相比都具有显著优势.

A cache strategy based on community nodes division and content popularity(CCNCP)was proposed to improve content availability due to the limit capacity of the routers in named-data networking(NDN).The regions were divided according to the topology structure,and the importance of nodes was comprehensively measured to select local central nodes and global central nodes.The contents with different popularities were reasonably cached in nodes with different importance.The classic GN algorithm was used to obtain partition communities.The centrality indexes of degree centrality,closeness centrality and betweenness centrality were defined.The system model was established,and the cache strategy was provided.To evaluate the performance of the scheme,the Icarus was used to simulate CCNCP and other alternative strategies.The simulation experiment was implemented on two topologies of Zachary karate club network and GARR of real topology.The results show that CCNCP performs better in both topologies.When the parameter S is equal to 0.25,the cache hit ratio,latency and path stretch ratio of CCNCP and the second best cache strategy ProbCache are 30.9%and 26.8%,52.9 and 56.2 ms,0.584 and 0.617,respectively,and the former is 4.1%higher,5.9%lower and 5.3%lower than the latter,which indicates that the proposed CCNCP has significant salient performance over the current typical strategies in all three metrics.