Researcc On File Caching Strategy Of Satellite CCN Network Based On Deep Learning

In the future,communication services need to achieve full coverage of the earth’s surface.However,the ground network cannot technically cover remote areas and extreme terrain.Therefore,the hybrid satellite ground network has higher reliability and coverage,and has become a solution to achieve global interconnection.However,the satellite ground network has scarce bandwidth and inevitable delay.Simultaneously,network traffic has shown explosive growth,in order to enable users to find interesting content in satellite networks faster,in-network caching is essential.As a novel network architecture,Conten t-Centric Networking(CCN)can meet today’s needs for content access and delivery.Therefore,in the future wireless communication,Satellite-integrated Content Centric Networking(SCCN)will be a promising candidate solution.However,the popularity of network files is constantly changing with time and geographical factors,and it is impossible for satellite nodes to cache the most popular files,resulting in low caching efficiency.Therefore,to address these problems,the Satellite-integrated Content Centric Networking(SCCN)is studied and a Deep Learning-enabled File Popularity-aware(DLFP)caching replacement strategy is proposed in order to achieve effective file distribution in SCCN.Aiming at the problem of file popularity prediction,a deep learning framework suitable for SCCN is proposed,called Deep SCCN,which can predict the final popularity based on the early request process of files.Deep SCCN models the complete request path of the file,uses recurrent neural network to train the input path information,and obtains the characteristics of the satellite network that really affect the popularity to predict the most accurate future popularity of files.By comparing with the classic feature-based popularity prediction model,the prediction results of the proposed Deep SCCN framework are more accurate in the case of different popularity distributions.Aiming at the problem of SCCN network topology changing,a Virtual Location Division(VLD)mechanism is proposed,which keeps the return path of content dat a packets unchanged by remapping the time-varying topology of the network to the static topology with virtual nodes.At the same time,in order to make CCN suitable for VLD mechanism,Forwarding Information Base(FIB)and Pending Interest Table(PIT)are improved,and the basic routing mechanism of SCCN is proposed to ensure the normal network running.In addition,a Minimum Delay File-caching Set(MDFS)algorithm is proposed at the end,combined with the prediction results of deep learning.MDFS enable nodes in SCCN to cache the most popular file set in real time.As experimental results shown,compared with several existing cache replacement strategies,the proposed method can significantly reduce the average access delay of users.