播放行为驱动的网络视频分发模型及相关算法研究

项目名称： 播放行为驱动的网络视频分发模型及相关算法研究

项目编号： No.61472455

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 刘宁

作者单位： 中山大学

项目金额： 83万元

中文摘要： 随着网络视频业务的飞速发展与普及，面对激增的用户和向高清发展的媒体内容，如何高效地进行视频分发，成为网络视频产业面临的关键问题。同时，海量视频用户的播放行为（如访问模式、视频热度、服务跳转模式等），对网络视频分发系统日益产生重大的影响与冲击。然而，由于缺乏对播放行为特征的理论分析和综合考虑，现有的网络视频分发服务无法采取有针对性的存储、调度策略，难以有效应对动态变化的播放环境。针对于此，在前期对某省网络视频系统播放行为分析与建模的基础上（190万用户连续150天超过60亿条播放记录），本项目将播放行为特征引入网络视频分发的建模与优化。首先研究用户播放特征预测算法，在此基础上，针对网络视频服务的不同系统角色，提出播放行为特征驱动的协同缓存、服务器选择和流量工程的联合建模与优化算法，提升系统动态服务性能，有效地应对海量用户动态行为对视频分发服务的挑战！

中文关键词： 用户行为；交互模式；流量工程；协同缓存；服务器选择

英文摘要： As the internet and multimedia technology are developing rapidly, network video service is becoming more and more popular. Facing massive users and higher definition videos, the key problem of the video service providers is how to effectively deliver the videos to users. At the same time, users' behaviors (such as access pattern, video popularity and switching mode) are becoming increasly impact on the quality of video service. However, the lack of thesis analysis of viewing behavior, current video delivery model and algorithm can't optimize the video storage and retrieval accordingly. And it's difficult for the video service system to self-adapt dynamically to the changing environment. In light of this, our former research has analyzed more than 6 billion records of 1.9 million users, and the future research will try to model the video delivery based the viewing behavior. Our project will study how to model and predict the users viewing behavior, and propose a joint optimal strategy for content providers and internet service providers. The joint optimal strategy based on the user's viewing behavior driven, uses the dynamic programming to achieve the best performance of cooperative caching, server selection and traffic engineering. Our method, which can optimize the data storage and reduce network traffic, expects to respond effectively to the challenge of massive users.

英文关键词： User Behavior;Interactive Mode;Traffic Engineering;Cooperative Caching;Server Selection