NDN中边缘计算与缓存的联合优化  

Joint optimization of edge computing and caching in NDN

作　　者：张宇[1,2] 程旻 ZHANG Yu;CHENG Min(School of Information and Electronics,Beijing Institute of Technology,Beijing 100081,China;Shanghai Institute of Mechanical and Electrical Engineering,Shanghai 201109,China)

机构地区：[1]北京理工大学信息与电子学院,北京100081 [2]上海机电工程研究所,上海201109

出　　处：《通信学报》2022年第8期164-175,共12页Journal on Communications

基　　金：国家重点研发计划基金资助项目(No.2019YFB1803200)。

摘　　要：命名数据网络(NDN)基于内容名称进行路由,且节点配备一定的缓存能力,故在架构上更易与边缘计算结合。首先,提出一个在NDN中实现网络、计算和缓存动态协调的综合框架。其次,针对不同区域内容流行度的差异性,提出基于矩阵分解的局部内容流行度预测算法;以最大化系统运营收益为目标,利用深度强化学习解决计算和缓存资源分配以及缓存放置策略的联合优化问题。最后,在ndnSIM中构建仿真环境,实验证明所提方案在提高缓存命中率、降低平均时延和远程服务器负载等方面具有明显优势。Named data networking(NDN)is architecturally easier to integrate with edge computing as its routing is based on content names and its nodes have caching capabilities.Firstly,an integrated framework was proposed for implement-ing dynamic coordination of networking,computing and caching in NDN.Then,considering the variability of content popularity in different regions,a matrix factorization-based algorithm was proposed to predict local content popularity,and deep reinforcement learning was used to solve the the problem of joint optimization for computing and caching re-source allocation and cache placement policy with the goal of maximizing system operating profit.Finally,the simulation environment was built in ndnSIM.The simulation results show that the proposed scheme has significant advantages in improving cache hit rate,reducing the average delay and the load on the remote servers.

关 键 词：命名数据网络 边缘计算 缓存策略 深度强化学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]