Design And Implementation Of Cache Strategy Based On Information-centric Networking

Information-Centric Networking(ICN)is a revolutionary network architecture.Named Data Networking(NDN),as its important representative,takes content as the center and routes it by name.The most important thing is to add the in-network cache function,which is an important design to improve network performance.In order to solve the problems existing in the current in-network cache,this paper proposes a novel deep learning-based caching strategy.First,by analyzing the characteristics of data(especially,the sensory data in the IoT scenario),we derive the rules of the user's request for data,and then discover its relationship with the cache.Then,based on the characteristics of the data and the simple unique characteristics of the data naming,we designed a new data/interest packet naming method,which combines the data characteristics into the name,provides powerful feature to support for the subsequent cache replacement algorithm and predict data popularity.And makes routing decisions based on the new naming method.Next,the deep learning-based popularity modeling method is the core of the algorithm,the popularity of the data is predicted,and the cache and replacement decision are made according to the ranking of the data popularity.We also design a two-level cache,adding the prefetch decision.Finally,the network topology is designed in the named data network forwarding daemon(NFD),and our strategy is implemented to verify.The results show that the deep learning-based caching strategy proposed in this paper can accurately predict the trend of popularity,greatly improve the hit rate,reduce the network delay,and break the bottleneck of the traditional cache replacement algorithm.