【大规模视频点播系统建模及优化理论研究】杨灿.pdf

ADissertationSubnitted to AcademicEvaluating Committee of HuazhongUniversity of Science andTechnologyfor theDegrccof Doctor of Philosophy in Engincering ModelingandOptimizationStudy onLarge-scaleVideo-on-DemandSystems Ph.D.Candidate:Yang Can Supervisor:Prof.

华中科技大学博士学位论文新的非均匀智能分段方法UESM,针对均匀的周期性广播分段方法ESM、金字塔广播PB和摩天大厦广播SB方案进行比较仿真研究,获得响应用户VCR请求的等待时间的理想结果和最佳的用户等待时间标准偏差、系统拒绝概率和信道吞啡能力等方面的优越性。UESM算法对历史信息进行数据挖掘,揭示用户对视频节目内容感兴趣程度的潜在分布规律,具有良好的自适应能力和学习智能,算法性能分析和模拟实验的结果说明该方法在实际应用中是可行且高效的.针对视频点播中VCR操作的业务特性,本文第六章提出-种基」MuIticast 视频显示同步控制(DRSC)策略。

华中科技大学博士学位论文 probability or that of sum of latency.By the experiments with MDP,the goodresultsin throughput and defection probability,which is the goal tOptimize,have bcen acquired.Besides,chaptcr 4 proposes afuzzy control scheduling policy called as FCSP,along with which the users class and cost