Research On Content-popularity Prediction And Caching Replacement Schemes In Smart Identifier Network

With the emergence in large numbers of new-pattern network applications,the existing TCP/IP architecture,which bases on the host centric communication mode,is difficult to realize the network information/content acquisition and transmission by proximity,resulting in low utilization of network resources and poor service quality and many other problems.The National Engineering Lab for Next Generation Internet Interconnection Device in Beijing Jiaotong University has proposed the Smart Identifier NETwork as a new clean-slate architectural.The Smart Identifier NETwork emphasizes the deployment of ubiquitous caching(In-network Caching),in which the cached router caches its forwarded content,and when the content is again requested,any cached router that caches the content can serve as a content source.This mechanism guarantees to get resources in the neighbourhood,reduces the consumption of network bandwidth and reduces the access delay.In this thesis,we design new caching replacement algorithms and mechanisms based on betweenness and predicted-popularity under the Smart Identifier Network from the perspective of improving the cache hit ratio,controlling the caching cost,and increasing the caching gain based on an overview of the existing caching mechanisms.First this thesis introduces the background of the Smart Identifier Network and the technology of the In-network Caching.Second,in this thesis,we describe an algorithm of popularity prediction based on Auto-Regressive(AR)model under the Smart Identifier NETwork,by summarizing and analyzing the statistical methods of popularity in traditional networks.This algorithm can predict the popularity of the content in the Smart Identifier NETwork in the next period accurately,and it can improve the uses of the feature of content popularity in the network.This algorithm provides a reliable and effective performance parameter to improve cache hit rate,reduce network traffic and reduce service acquisition delay.Third,this thesis proposes a new cache replacement schemes based on betweenness and predicted-popularity,for the sake of making the content router in the Smart Identifier NETwork cache more valuable service content.This schemes considers the content predicted-popularity value mentioned above,and the betweenness which can represent the network topology information.This schemes designs a caching replacement mechanism in which the content routers can determine whether to trigger the caching replacement behavior or not by the results computing by itself.The simulation results show that this cache replacement schemes based on betweenness and predicted-popularity can reduce the redundancy of network nodes,improve the cache hit rate and reduce the cache hit distance.This schemes,which has a good scalability,can also be deployed in the caching strategy for the purpose of predicting users' future requests.At last,the fifth chapter gives a summary of the whole thesis and makes some suggestions for the further work.