Research On The Problem Of First Miss In Edge Cache

With the rapid development of network technology,ubiquitous mobile devices and a large number of emerging applications lead to the explosive growth of mobile data traffic.Content delivery network(CDN)is an important means to solve the problem of rapid growth of network traffic.In order to get lower latency and better user experience,its caching devices are constantly sinking from the central network to the edge network.We call this kind of CDN edge caching network.The new scene brings new problems.In the edge cache,the cache device is usually mounted on the base station,which leads to a significant reduction in the number of users that can be covered by the cache,so that they receive relatively few requests.Based on the analysis of the YouTube request data set,we find that the global hot content in the edge cache is not necessarily hot content in the local cache.Meanwhile,the content in the local cache in the edge cache is less hot,and a large number of content is only visited a few times.We call the first request miss of a certain content received by the caching device as the first miss.Through these analysis and literature search,we find that the first miss has an important impact on the overall hit rate of the cache in edge caching,which is a unique and neglected problem in edge caching.As far as I know,this thesis is the first to raise this question.This thesis adopts the strategy of active push to solve this problem.Different from other articles,we do not push content according to the heat of content,because we find that there is a small difference in the heat of content in the edge cache.We use the idea of collaborative filtering for reference and push the content through the similarity between caches.We first propose the simplest broadcast strategy,and prove that the broadcast strategy can solve the first miss problem in the edge cache under the condition of unlim-ited cache.Then we calculate the Jaccard similarity between the caches according to the request set received in the cache history,limit the scope of content push by clustering the cache,and use the graph segmentation algorithm to solve the clustering problem Further-more,it is shown that the problem has submodule property.Finally,the probability of content push is increased based on clustering to further improve the accuracy of content push.We use SNM model and Youtube statistical data set to generate two request se-quences,and simulate the similarity of cache preferences.In the simulation experiment,we compare the performance of the proposed content push strategy with that of the cache strategy proposed in recent studies,including cache hit rate,cache push overhead,and cache push accuracy.It is verified that the proposed content push strategy can solve the first miss problem in edge cache to a great extent with acceptable overhead.