Design And Implement Of Large-scale Streaming Media CDN

CDN (Content Delivery Network): A system of distributed content on a large intranet or the public Internet in which copies of content are replicated and cached throughout the network. When content is replicated throughout the country, or throughout the world, users can have quicker access to it quickly from "the nearest and the best" site. This paper mainly studies on video steaming CDN system architecture, steaming media user pattern and a distribution technology based on "load and hot-degree" of streaming media.First of all, this paper designs the architecture of streaming media CDN, mainly organizing in tree topology. Combining the partial caching model with the storing the whole replica model, it implements multilayer and multigrain data pattern by replicating, caching(original server to surrogate) and buffering(disk to memory). Scheduling algorithm automatically adjusts structure-grain through tracing change of load-distribution and content popularity. Replicating and caching provides closely accessing for users, and aggregates flow of users accessing at PoP, and makes flux localize. Cluster nodes cooperative sharing buffers aggregates access to disks, alleviates disk I/O bottleneck.Next, through analyzing nearly seventy million rows log data of four media servers of stvod. com, we gets ten conclusions for characteristics of streaming media and its access pattern. These conclusions give evidences for designing and improving algorithms of content delivery and load balance in streaming system.At last, according to analyzed conclusions of streaming media characteristics and its access pattern, this paper proposes a streaming media delivery technology based on "load and hot-degree". Adopting events triggering mechanism and periodically monitoring mechanism, it traces changes of objects' hot-degree and nodes' load, and automatically decides to add, delete, migrate or degrade the replica to adjust quantity of replica, position of replica and structure-grain. It can resolve the problem of the skewed-load, which is aroused by unbalance of data load distributing. Running of the system indicates that this streaming media delivery based on "load and hot-degree" is a valid replica dynamic management mechanism.