Video Transmission Optimization In Software Defined Network

Streaming media services enrich people's lives,but also incur the explosive growth of Internet traffic.Coupled with the complexity of user behavior,a high redundancy rate is created.Fortunately,the popularity distribution of video traffic conforms to the 80/20 principle,and in-network cache is an effective mechanism for eliminating repetitive traffic by caching popular content in network nodes.Software-defined network(SDN)enables centralized control and management,facilitating collaboration between network layer and application layer.Inspired by this,we propose an SDN assisted multimedia streaming VoD system integrating in-network cache,which uses the caching capabilities of network nodes(OpenFlow switches)to cache highly-popular content,providing near-by services at the edge of the network while reducing the consumption of core network bandwidth and server load,and improves the quality of service(QoS).With the separation of SDN's centralized management and control forwarding,the network state can be perceived in real time,and video streaming and cache management can be implemented according to the deployed cache strategy.In order to effectively utilize the limited storage resources to improve the cache hit rate,this thesis also introduces the Network Functional Virtualization(NFV)technology to support in-network cache and video streaming functions.Considering the caching ability of network nodes,traditional caching algorithms for full file granularity can not make full use of limited caching resources,and Interval Cache technology can support video clip caching and permutation.Therefore,this thesis designs an intranet caching strategy for RTP streams to support variable-length interval caching.By aggregating content requests from multiple users,cache reuse can be achieved.The length of cached video clips can be adaptively adjusted according to different access modes of users,and the utilization rate of cache resources can be improved.In addition,the cache state can be maintained conveniently by setting up a state machine in the cache window,and the cache can be replaced actively according to the trend of the popularity of the cache content,thus improving the cache efficiency.In the design architecture,the cache function is abstracted into a cache policy module separately,which increases flexibility and improves performance by deploying specific scheduling policies.In order to verify the feasibility and practicability of the above scheme,we design and build a prototype system,define the relevant performance metrics,and use three control algorithms to conduct comparative simulation experiments,through simulation experiments to evaluate the performance improvement can be achieved.