Research On Video Caching Strategy In Mobile Edge Computing Networks

In recent years,with the development of mobile communication technology,the number of mobile devices in the wireless communication network has grown explosively,and a large number of new Internet applications have emerged,especially the rapid development of mobile video service,which leads to the increasing proportion of the traffic generated by video data in the communication network.In the scenario where high bandwidth consumption video files are used as the service access hotspot,the traditional cloud computing mode is limited by the bandwidth capacity of the backhaul link,and it is difficult to respond to large-scale video requests at the same time,resulting in frequent network congestion and long response delay.In order to overcome the limitations of the existing network architecture,researchers proposed the Mobile Edge Computing(MEC)architecture,which provides Computing and storage resources near the network Edge,in order to relieve the traffic pressure of the backhaul link and reduce the response delay of the request.In the video business,it is obvious that using MEC server to cache video resources can effectively deal with the problems caused by the rapid growth of mobile video traffic.However,considering the limited cost investment,marginal resources cannot be arbitrarily deployed to meet the growing needs of users,so the performance of caching strategy is particularly important.Existing caching strategies based on LRU and LFU assume that users' access mode is fixed,that is,recently requested or frequently requested videos have a higher probability of being accessed in the future,but in mobile video networks,popular content always expires soon,and the past access mode often cannot accurately represent the popularity of videos in the future.Some studies have tried to select user preferences,request records and other information to model the popularity.However,in addition to explicit features,some implicit features in mobile video network,such as user mobility,user interaction behavior and network state,will also have an impact on the popularity of video,resulting in the difficulty of accurate modeling the popularity of such strategies.In order to improve the performance of the cache strategy under dynamic changing scenarios,some studies proposed a cache strategy based on deep reinforcement learning.In the absence of any prior information about the popularity of content,this kind of strategy uses deep reinforcement learning technology to perceive the state of the environment and autonomously learns the cache strategy according to the observed environmental data.But most of these schemes do not consider the choice of video bit rate.In addition,most of the existing studies take hit rate and traffic offload as performance evaluation indexes,leading to the failure to fully consider the impact of caching strategy on the quality of user experience when selecting bit rate.Considering the edge mobile computing under the network environment,network status of degeneration,user mobility,the characteristics of video files will affect caching strategies,such as how to efficiently use the edge nodes and wireless network with limited resources,to provide users with high quality and low latency of video service is an important topic of edge of mobile computing.This paper focuses on the research of cache update and placement strategy for video services under mobile edge computing network.The main work is as follows:(1)Video cache update strategy based on adaptive bit rate selection in mobile edge computing networkAiming at the problem that it is difficult to accurately model video popularity and network state in mobile edge computing network,a video cache update strategy based on adaptive bit rate selection in VOD service is studied.Firstly,a two-layer video cache architecture based on mobile edge computing network is proposed.Under this architecture,the MEC platform has the capability of caching,transcoding and state awareness of wireless network.The cache system can design the cache strategy by using the functions provided by the MEC platform.Secondly,a cache update optimization problem is defined to maximize the Qo E revenue of the system and users.Finally,a deep reinforcement learning based cache update and bit rate selection algorithm is designed,which takes the video history request information and the user history network information as the environment state.The algorithm can perceive the state information of the environment and learn the caching strategy independently according to the feedback of the environment.The proposed scheme also optimizes the selection of cache update time,cache video and its bit rate.The simulation results show that the scheme proposed in this paper can effectively improve the system cache income and user experience quality.(2)A multi-base station cooperative short video cache placement strategy with user location awareness in mobile edge computing networksBecause the popularity of short videos changes rapidly and the request probability is closely related to the user location distribution,a collaborative caching strategy for short videos considering user mobility is studied in mobile edge computing networks.Firstly,a cloud edge collaboration architecture under mobile edge computing network is proposed.This architecture is equipped with an acer station and a micro-base station.Each micro-base station can communicate with the acer station,and the neighboring edge nodes can cooperate with each other in caching.The recommendation system of the short video platform generates the short video set to be pushed according to the user's preference and determines the popularity ranking of the short video.Before caching the video,the edge node uses the mobility management module to predict the base station where the user is in the next time slot,and the cache control module decides the location of the short video cache according to the user distribution.Secondly,the user mobility and cache placement problems are modeled,and the cache placement optimization problem is defined to maximize the revenue per unit traffic and delay.Finally,based on the LSTM network model and deep reinforcement learning,a cache placement algorithm with user location awareness is designed.The algorithm can accurately predict the movement trajectory of users,and cache short videos to appropriate locations according to the user's movement,so as to meet the short video requests of more users.The experimental results show that the proposed scheme can effectively reduce the flow of the backhaul link and reduce the delay of video transmission.Further experimental testing and analysis show that it is very important to accurately predict the target location of the user's movement and select the appropriate location of the short video to improve the cache performance.