Research On Cache Consistency In Named Data Networking

The major usage of the current Internet has shifted from the information browsing to the content dissemination.However,the traditional TCP/IP network based on the host-centric architecture is hard to cope this change.Thus,a data-centric architecture of the network,Named Data Networking(NDN),is proposed by the researchers as a candidate architecture of the future network.One of the typical characteristics of NDN is ubiquitous caching,which on the one hand enhances the data transmission efficiency but on the other hand raises a problem of cache consistency.Guaranteeing the cache consistency is to ensure the content copy acquired by the user from the cache is valid.It has great significance for NDN to promote service capability and broaden application scope.In addition,when providing the cache consistency,the cache system also brings the extra overheads on management and traffic of the network.How to reduce these overheads without affecting network performance is also a problem that needs to be solved when deploying consistent services in NDN.To address the above problems,in this work,we first establish a mathematical model of the cache consistency,based on which the influence factors of network perfor-mance in the basic cache consistency scenarios is able to be analyzed theoretically and some common insights of the design of cache consistency strategy are also obtained.The specific research work is as follows:(1)By using Poisson process and conditional probability to character the traffic be-havior and the interactive relationship between existence and validity respectively,we develop an analytical model of four different invalidation schemes with LRU replace-ment.This model has a extremely high accuracy in predicting the cache hit ratio and the server load under arbitrary update frequency distribution.(2)According to the developed model,we propose a cache consistency strategy of proactive invalidation with optional renewing,called PIOR,to provide strong cache consistency for NDN.PIOR leverages the lightweight publish/subscribe model of NDN,actively publishing the updated content to the cache nodes where the updated content copies have been stored before.Several publish/subscribe rules are also set in PIOR to reduce the extra overheads.In order to evaluate the performance of the proposed model and strategy,we con-duct extensive simulation experiments.The experimental results show that our model can achieve a higher prediction accuracy and better extensibility against other models and PIOR has a higher cache hit ratio,lower server loads as well as a smaller propor-tion of consistency traffic against other strategies when providing cache consistency for NDN.