摘要： 云模式下的流媒体服务系统需要有效与平滑的资源调度。传统的last-minute迁移大多只将负载信息用于迁移策略计算,但当流行度动态波动时,间接且单一的方法无法做出准确的策略调整。为此,将流行度分布应用于迁移模型中,提出一种基于动态阈值分配的会话迁移策略,依据流行度分布,确定每台服务器上各类视频的会话分配阈值,通过分配阈值的指导性避免请求接入及会话迁移的盲目性。实验结果表明,该策略能有效改善请求接受率,并保持较低的迁移代价。

关键词: 流媒体边缘云, 资源调度, 会话迁移, 流行度分布, 动态阈值分配, 迁移代价

Abstract: For the streaming media system based on the cloud model,more efficient and smooth resource scheduling is strongly required.In the traditional “last-minute migration”,only the load information is used for the migration scheduling.Due to the dynamic fluctuation of the video population,such indirect and single way cannot guarantee effective scheduling.Besides the load information,the video popularity is also used for the migration model,this paper propose a dynamic threshold allocated migration strategy.The allocation thresholds of different videos on each server are obtained on the basis of the new video popularity,and due to the guidance of the allocation threshold distribution,the blindness of request access and session migration can be avoided effectively.Experimental results show that the strategy cannot only effectively improve the request acceptance ratie,but also keep the migration cost low enough.

Key words: Streaming Media Edge Cloud(MEC), resource scheduling, session migration, distribution of popularity, dynamic threshold allocation, migration cost

中图分类号: 

• TP391