A Distributed Probabilistic Mobility-Aware Caching In Small-Cell Networks

With the rapid expansion of mobile traffic,a huge challenge is brought into the design of new network architectures.Facing the pre-cache deployment in small-cell network,which is one technology of 5th generation wireless system(5G),we propose a mobility-aware caching system.It makes the requested file closer to users through the optimal deployment,thereby reducing the download delay and reducing the redundant data in backhaul.The core idea is to establish a mobile user single-step transition model with big data,combine the sojourn-time patterns to achieve the time-related trajectory prediction,and then encode the requested file into time-series segments according to the download rate.The main methods are listed as follows.1.Two schemes based on mixed Markov model and Conditional Variational Autoencoders are respectively adopted to implement single-step transition model.2.Power law is utilized to fit the sojourn rules to com-plete the time-related trajectory prediction.3.RaptorQ is utilized to encode the file into multiple segments,which have one-to-one correspondence with time series,in order to calculate the optimal storage location in rank for each segment.4.Considering the restricted condition about the cache space of base station,we formulate the caching problem into an optimization framework with the objective of minimizing the cost of downloading,and solve it by Greedy Algorithm.In the experiment,firstly,we conducted the trace-based simulation to verify the feasibility of both prediction methods and analyzed their performance.Secondly,we used a large number of sample data to analyze the sojourn rules and got satisfactory results.Finally,we conducted the simulation of pre-cache deployment and compared with some other cache strategies.As a result,it is proved that the proposed system can greatly improve the hit rate of pre-cached file and achieve pre-cached load balancing.