Research On Edge Caching Technology Based On Information Timeliness

With the development of mobile communication technology,the number of smart terminal devices has exploded.With the continuous emergence of new applications,users have begun to subscribe to various types of real-time services,such as social networks,live video,online games,weather monitoring,and so on.In this kind of content-oriented services,the requirements for delivery delays are becoming more stringent.In the traditional centralized architecture,requests for contents are usually forwarded to the content server of the core network.Returning data to users leads to congestion of the backhaul link,which in turn increases the delay in obtaining content and reduces the quality of service(QoS).Therefore,making full use of cache resources at the base station and users' equipments can achieve efficient content delivery.In traditional wireless edge network cache strategies,delay,throughput,hit rate,or energy efficiency are often used as indicators to measure the performance of cache strategies.In the new real-time application scenarios,the content is time-sensitive,and the aging content may even be misleading to users.Therefore,how to design a caching strategy with content timeliness as a performance indicator is a question worthy of study.Based on the concept of information age,the thesis studies the strategy of content delivery and caching placement based on optimal timeliness.The main contributions of this article are as follows:(1)An energy-efficient content access strategy with content timeliness constraints is proposed under the D2D(device-to-device)-assisted edge caching architecture.Due to the short communication distance,compared with the traditional way of serving users by cellular link,serving users through D2D link has higher energy efficiency.However,since the updates of users'cache are mostly passive,the real-time content cannot be guaranteed,so the timeliness of the content obtained through D2D link is low.The thesis takes the timeliness of contents as a constraint and models the problem of minimizing the long-term energy consumption of the entire network as a Markov Decision Process(MDP).The thesis use the Bellman equation and the inherent relationship between the timeliness of the content and the energy consumption under the original constraints to transform the problem of minimizing long-term energy consumption into a discrete time-slot problem.Finally,simulations verify the superior performance of the proposed strategy in terms of energy consumption and timeliness.(2)The caching architecture is further extended to highly dynamic vehicle networking.Vehicles can obtain content from the BS(base station)or RSU(Road Side Unit)by V2I(Vehicle-to-Infrastructure)links or from surrounding vehicles by V2V(Vehicle-to-Vehicle)link.It's more efficient to serve users by RSU or vehicles.The thesis defines the age value function to represent the impact of content at different ages on vehicles due to the timeliness requirement,and defines the age value weighted hit rate as the effective hit rate.The thesis proposes a V2V collaborative caching placement strategy to maximize the effective hit rate.The dynamic characteristic of vehlcular networks brings two problems:the limited coverage of the RSU and the Spatiotemporal variation characteristic of content popularity distribution.The thesis firstly uses a cache assistant selection algorithm based on trajectory prediction.Vehicle nodes that have a higher probability of reaching hotspots are chosen.The thesis proposes a content-popularity prediction algorithm based on Multi-Armed Bandit(MAB).Finally,the content placement problem that maximizes the effective hit rate is modeled as an integer programming problem and is solved by a maximum effective hit rate incremental algorithm based on the greedy algorithm.The effectiveness of this strategy in terms of hit rate and timeliness is proved by simulation.