论文部分内容阅读
在VOD(Video-On-Demand)系统中,由多服务器组成的机群比单一的VOD服务器具有更好的可扩展性,同时异构机群比同构机群有更好的实用性、灵活性,也面临更复杂的负载平衡问题.在多VOD服务器异构机群系统中,数据分布策略非常关键,它对整个系统的请求调度、负载平衡、可扩展性等有很大的影响.本文针对数据分布问题,结合有关客户访问模式的统计信息,来确定热门影片的访问概率,再以数据复制的方式,根据影片热门度的不同和各异构服务器上的资源情况,将文件分布到机群中服务器上.实验表明,通过给出的针对异构机群的副本产生和放置模型,提高了客户可扩展性并达到较低的负载不平衡度.本文的策略将客户的行为方式和服务器端组织框架结合起来,既解决了客户的可扩展性问题,又实现了机群内的负载平衡.
In VOD (Video-On-Demand) system, a cluster composed of multiple servers has better scalability than a single VOD server. Meanwhile, heterogeneous clusters have better practicability and flexibility than homogeneous clusters. Facing more complex load balancing problem.In multi-VOD server heterogeneous cluster system, the data distribution strategy is very critical, which has a great impact on the entire system’s request scheduling, load balancing, scalability, etc. In this paper, the data distribution , Combined with statistical information about customer access patterns, to determine the popularity of access to popular movies, and then the way of data replication, according to the popularity of the film and the heterogeneous server resources, the files are distributed to servers in the cluster. Experiments show that, by giving a replica generation and placement model for heterogeneous clusters, it improves customer scalability and achieves lower load imbalance.The strategy of this paper combines customer behavior and server-side organizational framework, Both to solve the problem of scalability of customers, but also to achieve load balancing within the cluster.