支持高并发度的大容量媒体库

Large-Capacity Media Library Supporting Highly Simultaneous Access

支持大容量和高并发度的媒体库越来越流行.利用传统的基于纯粹的服务器/客户端、对等网络或网格的方法构造这种类型的媒体库,难以同时满足高并发度和可靠服务的要求.提出了一种结合了网格和对等网络(peer-to-peer)的媒体库架构——NeoMedia.专用的和志愿参与服务的节点在系统服务器的支持下,形成了一个大容量的虚拟媒体存储池,请求服务的节点在从媒体存储池下载媒体文件的同时,在系统服务器的协调下,采用对等网络的方式相互提供服务,从而进一步提升系统的性能.系统服务器还根据虚拟媒体池的负载模式和强度的变化自动调整服务的资源分配,自适应地优化系统的性能.NeoMedia面向的是下载完毕再观看这种可以充分利用服务器的带宽和客户机能力的模式.理论分析表明:系统能够在使用较少的系统带宽的情况下,支持巨大的并发的用户请求,并且提供非平凡服务.

Large-Capacity media libraries supporting highly simultaneous access are more and more popular. Media libraries based on traditional client/server mode, or pure Grid or P2P can hardly meet the requirements of both high concurrency and reliable service. This paper proposes a new architecture combining Grid and P2P for media library-NeoMedia. It consists of the dedicated and volunteered nodes supported by the system server and forms a virtual media pool of large capacity. Coordinated by the system server, nodes requesting files from the media pool serve each other in the style of P2P, which further enhances the system＇s performance. According to access pattern and intensity, the system server automatically adjusts the allocation of resources, adaptively optimizing the system＇s overall performance. NeoMedia targets at view-after-downloading, which can efficiently utilize servers＇ precious bandwidth and clients＇ power. Theoretical analysis demonstrates that NeoMedia is able to support hugely simultaneous requests and provide non-trivial services at the same time with relatively low server-side bandwidth consumption.