Measurement Study And Edge Cache Strategy Research On Social Instant Video Platform

With the popularity of smart phones and mobile networks,social instant video platforms are gradually emerging,which bring a lot of traffic to the backhaul link of mobile network.Compared to VoD platforms such as YouTube,the distribution of video popularity in social instant video platforms tends to be much more skewed,so using edge caching technology,caching popular videos in advance on the base station can greatly reduce users' latency in watching video and the traffic load on the backhaul link and cloud servers.But video popularity in social instant video platforms changes rapidly and is more dynamic,so the traditional cache replacement strategy based on the user's recent visit record is difficult to respond to changes in video popularity in a timely manner,resulting in reduced caching efficiency.In this dissertation,we first obtain the API of Douyin by reverse engineering and use the obtained API to crawl Douyin's data.Using Douyin as a case,this disserta-tion characterizes the users and videos and measures the video propagation mechanism in detail.The measurement shows that the distribution of users in Douyin is uneven,with numerous users clustered in a few areas within the city area.The main method of video propagation in Douyin is recommendations,followed by social networking,and the video propagation exhibits significant temporal localization and spatial local-ization.Based on the measurement results,this dissertation proposes a collaborative edge caching strategy with cloud-side collaboration to reduce the load on backhaul link and cloud servers and users' latency in watch video.The uneven distribution of users re-sults in unbalanced load of base stations in different regions,so the algorithm proposed in this dissertation considers the edge nodes in the region collectively,enabling the edge nodes with low load to serve the edge nodes with high load and all nodes collaborate with each other to improve resource utilization rate.Measurement shows that Douyin is able to largely perceive the propagation of video,so this dissertation adopts the strategy of cloud-edge collaboration,which uses cloud servers to predict which videos users will watch and edge nodes to cache videos.Finally,this dissertation uses the crawled data to conduct a simulation and test the strategy,which achieves a better result in terms of cache hit rate,traffic offload rate and latency reduction rate comparing with the tradi-tional cache replacement algorithm.