Research On Optimization Of Multi-stream Network Video Service System

With Internet infrastructure developing rapidly,PCs has been gradually replaced by emerging terminals such as smartphones,tables and smart TVs which have become main devices for video playback.Due to the large difference in screen sizes of these terminal devices,video service system needs to provide video files with multiple formats and codestreams,which puts great pressure on the cache space.The introduction of real-time transcoding technology will alleviate this pressure.This paper proposes a multi-stream video service system based on CDN(Content Delivery Network)framework.The system is mainly based on cache,real-time transcoding as a supplement,the vast majority of users request will be solved nearby.Thus,the transmission path of video stream can be shortened,the transmission pressure of backbone network can be alleviated,and the system's service capability and quality can be guaranteed while meeting the demands of large-scale video on demand in a user colony.The research content of this study mainly includes the following aspects:(1)Construct multi-stream video transcoding cost tree.Owing to that the problem that video provider's cache space is insufficient due to various screen sizes of terminal devices,a transcoding time overhead directed graph is established.Combining with the cache overhead,the transcoding time cost of different transcoding input files is analyzed,and the multi-stream video transcoding cost tree is constructed.Based on this,the paper studies the selection strategy of transcoding sources file before transcoding tasks.(2)Optimize system overhead.Storage overhead generated by video file cache and measurement standards of transcoding overhead generated in the process of video transcoding is unified.The overhead function is constructed by analyzing the storage overhead and transcoding overhead.Based on the actual transcoding capability of the system and the "8-second law" as the delay standard,the system overhead function is optimized by combining the real-time transcoding technology with the server cache strategy.(3)Simulate the large-scale video service system.Because it is difficult to build the experimental platform of large-scale network video service system,AnyLogic is used to model the large-scale network video service system,and the transcoding cost tree and system overhead function are applied to the model of network video service system.The whole process of interaction between users and video service system is simulated by running simulation model.The data of user waiting time is collected to analyze the user waiting time distribution before and after adding real-time transcoding technology.Experimental results show that the video transcoding cost tree constructed in this papercan effectively shorten the transcoding time of video.Therefore,deploying a cache system with transcoding services in a user colony can effectively improve the user experience.