Study Of Caching Strategies In Industrial Edge Networks

With the rapid development of Io T technology,the number of smart devices in industrial production sites has increased sharply,thus generating massive amounts of data and placing higher demands on data storage,processing and analysis.By deploying edge nodes at the edge of the network,edge computing provides storage and computing services close to the user,greatly reducing time overhead.The emerged edge computing is a good solution to fulfill the real-time requirements of industrial applications.Edge caching caches part of the content into the edge server,and the cache performance will greatly affect the response latency of user requests.In this paper,we study the cache optimization problem in edge networks for industrial application requirements,and propose corresponding cache replacement algorithms and cache prefetching algorithms.The main research elements are as follows.Firstly,the network structure in a typical industrial application scenario is summarised and a three-layer edge network model is constructed.Subsequently,based on the analysis of the existing cache problem optimisation objectives,the cache hit rate and the average transmission delay are taken as the optimisation objectives to establish the edge cache problem model.The shot noise model is also proposed as the industrial user content request model,and the analysis of simulation experimental results shows that the shot noise model can better simulate the dynamic characteristics of industrial data requests compared with the static model.Secondly,the characteristics of content and content requests in a typical industrial edge network are analyzed,and then a new cache replacement algorithm is proposed.It predicts the popularity trend based on the frequency of the attribute features of the requested contents in last periodic time window.It then defines the value of each cache content together with the popularity prediction,size and time updates parameters.The experimental results show that our proposed algorithm outperforms the five baseline algorithms,,which are most popular content(MPC),greedy dual size(GDS),least recent used(LRU),least frequently used(LFU),and FIFO algorithm.The proposed algorithm obtains the best performance of hit rate and average delay in all the testing cases with different parameter settings on user request models,file size distributions,and file types.Finally,a cache prefetching algorithm based on frequent sequence mining is proposed.The algorithm divides the multiple long sequences into short sequences with a cutting window size.And then It filters out strong association rules by finding the sets of high-frequency items under the constraint of sliding window.The association rules are matched with the new coming user request,and caches the content set that are associated.The proposed cache strategy could make better edge caching decisions by effectively predicting the time-varying user requests.The simulation results show that,compared with the baseline algorithm C-Miner,the proposed cache strategy improves the cache hit rate and average transmission delay.