Cloud Services Delivery Networks (CSDN) constructs a layer distributed server overlay over the Internet, which uses the way to the nearest and on-demand approach providing services to end users. Facing the scale and diversification of the resource demand characteristics of the Internet cloud services, CSDN forms different logical sub-server overlay for different kinds of cloud services. However, most servers and bandwidth resources of CSDN are used to deliver the streaming and downloading kind of cloud services, and the dynamic allocation of their delivery resource is the main research emphasis in this paper. This paper first models the problem to be a multi-dimensional facility location problem, according to the two characteristics; the memory resource and bandwidth resource of this kind of application are the bottleneck resource; the hot contents of this kind of application can be delivered using the Peer-to-Peer mechanisms. After the model analyzed and its NP-Complete proved, we then propose a heuristic algorithm. Finally, using the service delivery cost savings as the performance metrics, while the actual system's operation trace is as the input, the effectiveness of the algorithm are comprehensively assessed.%云服务传递网络(Cloud Services Delivery Networks,CSDN)在Internet之上构建了一层分布式服务器网络,以就近和按需的方式向用户提供云传递服务.面对互联网规模化和多样化云服务的资源需求特点,CSDN形成了针对不同类型云服务传递的逻辑子服务器网络.CSDN的很大一部分服务器和带宽资源用于流媒体和下载类云服务的传递,该类型云服务传递资源的动态分配问题是该文的研究重点.根据该类型业务内存资源和带宽资源同为瓶颈资源以及该类型热点内容可采用P2P机制的两个特点,文中首先将该问题建模为多维设备选址模型.然后在对该建模分析及其NP完全性证明后,提出了一种启发式模型求解算法.最后以服务传递开销节省作为性能评价指标,以实际系统的运行数据为输入,全面评估了该模型求解算法的有效性.
展开▼
