面向海量流媒体信道资源分配快速Nash议价算法
Fast Nash bargaining algorithm for resource scheduling problems with a large number of media streaming channels

在大规模在线流媒体分发系统中，服务端需处理来自全球各区域的海量用户请求。现有混合云架构不能很好地满足日益增加的动态流媒体内容分发要求，需结合私有数据中心、云和内容分发网络3类平台，充分挖掘各平台的优势以降低费用并提高服务质量。针对基于3种平台的混合云，该文给出了多资源分配问题的描述，将其转化为Nash议价问题，从几何角度获取问题的高效求解算法，并基于实际商用环境中海量流媒体采样数据进行了模拟实验。实验结果表明：相比传统的混合云架构，该算法可显著提升服务质量，在动态和静态内容混合情况下可降低平均约40%的费用，可在包括大量动态流媒体内容场景中进行快速有效的资源分配。
Abstract：The servers in large media streaming systems need to handle a large number of requests from all around the world. However, due to the increasing dynamic media content and because existing cloud-based architectures cannot provide enough benefits, the service provider needs to utilize a hybrid architecture composed of a content delivery network with private and cloud data centers to provide sufficient quality of service while reducing costs. This paper describes a general resource scheduling problem for this scenario for a hybrid cloud, which is then transformed into a Nash bargaining problem. A fast Nash bargaining algorithm is given based on a geometrical perspective of the problem. Tests show that the algorithm improves the quality of service and reduces expenses by about 40% compared with a traditional hybrid architecture, so it can effectively handle large amounts of dynamic media content.