Design And Implementation Of Streaming Media Publishing System Based On Collaborative Cache Management Strategy

With the continuous improvement of people's living standards,the Internet has become an indispensable part of life.With the gradual improvement of internet infrastructure and video coding and decoding technology,streaming media technology has developed rapidly.Audio and video streams account for a large proportion of Internet traffic.Streaming media technology has been widely used in all walks of life.In China,many internet companies started to provide streaming media on demand and live broadcast services.However,the emergence of a large number of online video streams has also put the Internet under increasing pressure.At the same time,users are increasingly demanding the performance,response time and user experience quality of video websites.Therefore,how to properly optimize the streaming media caching strategy to improve the quality of user experience and reduce the pressure on the server and backbone network becomes more important.The evaluation of video popularity provides a basis for cache server decisions,and real-time data-driven optimization and refinement of cache strategies.So this paper summarizes the problems faced by streaming media caches,researches popularity prediction,cooperative caching strategies,and data-driven E2 optimization framework.The main work is as follows:(1)The related theories of streaming media and the related technologies involved in the streaming media publishing system are briefly described.Focused on research and introduction of Apple's HTTP Live Streaming(HLS)system,including content preparation,content distribution,and client implementation principles and methods;(2)Research on Cache Management Technology in Current Streaming Media Publishing System.For single-node server cache management,the popularity indexing algorithm(HIPP)based on historical access record index weighting is researched and tested,and the implementation method of cache allocation based on HIPP algorithm is given.A multi-node collaborative cache management strategy(CCM)based on popularity is studied and experiments and tests are conducted on the algorithm.(3)In order to further improve the system,and solve the problems that may exist,a data-driven E2 framework(exploration and exploitation,E2)for improving the user experience was studied,and the algorithms for implementing the E2 framework were tested and verified.(4)Finally,this paper uses Ubuntu to build a video on demand system based on HLS protocol.The system uses Nginx as a video publishing server,uses FFmpeg to generate video slice and index files,and implements HIPP algorithm,HIPP-based cache allocation algorithm,cooperative cache management strategy and E2 framework in Python and Java.The test results show that HIPP algorithm,popularity-based cache allocation algorithm,cooperative cache management strategy(CCM)and E2 framework are feasible and effective.