Research On The Key Technologies Of Cost-Effective Video Delivery

With the unprecedented development of high-speed Internet and streaming media technology,Online video service has quickly risen to become a mainstream "killer"application for network bandwidth consumption.According to Cisco white paper,IP video traffic will be 82 percent of all consumer Internet traffic by 2020.In video Internet era,on one hand,the online video service extremely consumes the storage and bandwidth resources.On the other hand,the user requirements for high quality of service(QoS)are more demanding.Therefore,this thesis focuses on how to provide high QoS video delivery at controllable cost.This thesis utilizes data-driven approach,cloud computing platform and Internet exchange point(IXP)structure to study the problems including content management,resource allocation and scheduling,and cooperation between content delivery network(CDN)and Internet service provider(ISP).Contributions of this thesis are summarized as follows:1.Proposing adaptive content management architecture design for user generated content(UGC)video delivery.This thesis proposes a framework for UGC video delivery by introducing proactive content management mechanism and coupling content replication and request routing modules.Based on the measurement of access characteristics of UGC videos,this thesis embeds result of measurement analysis into name resolution system to drive adaptive content management.Simulation based on real trace has proved the validity of the architecture.2.Proposing content replication and request routing strategies for the delivery of UGC videos.This thesis reveals access characteristics of UGC videos which are closely related to content replication,based on data-driven method.Videos' popularity prediction framework is designed for guiding content placement and eviction.Meanwhile,cooperative request routing strategy which is content position aware is proposed.The strategies can improve hit ratio of edge servers,reduce average transmission distance and origin server's load.3.Studying resource allocation and request routing problems in video delivery network.This thesis models resource allocation problem to optimize the trade-offbetween leasing cost and QoS.It has been proved to be a NP-hard problem,and a heuristic greedy algorithm is designed for solving it.To adjust to the dynamic changes of users' requests over time,this thesis proposes an efficient online algorithm for request routing,aiming at optimizing user experience.4.Studying joint request and response routing problem in video delivery network.To minimize operational cost and enhance user experience,this thesis models joint request and response routing problem in multiple transit ISPs environment.The joint problem is decomposed into sub-problems which can be partially optimized by each access ISP via introducing Lagrange multiplier and converting to a dual one,thus can be solved in a distributed manner.The experiment demonstrates that the relative error between the proposed distributed algorithm with 50 iterations and theoretical optimum is around 0.7%.