云数据中心网络的QoS保障方法研究

云计算是当今IT业发展的热点，数据中心网络是云计算的核心基础设施，也是云计算的性能瓶颈。本项目分别针对公有云数据中心和私有云数据中心的特点和需求，研究其数据中心网络的优化方法以提供云计算的QoS保障：在私有云中提供流级别的QoS保障，在网络侧提出基于路径隔离与动态路径划分的方法，同时满足大流的吞吐量与小流的延时要求，在主机端提出支持多传输协议并存的新型协议栈架构，该架构可以根据应用程序的QoS需求对传输协议进行灵活配置与重组；在公有云中提供租户级别的QoS保障，设计基于多路径的带宽保障方法，提出多令牌桶技术进行多路径的限速以及动态速率调整。本项目的研究工作将为云数据中心网络的设计奠定必要的理论基础，优化后的云数据中心网络可以满足网络应用程序以及云租户的QoS 需求，优化用户体验，充分提高数据中心网络的利用率，降低云计算平台运营成本。

Cloud computing is a hot issue in the development of IT industry. As the infrastructure of cloud computing, network is the core component, but also the bottleneck of cloud computing. This project is aiming to meet different QoS requirements through the optimization of the data center network, according to the characteristics of both the public and private cloud data center network: In private cloud, QoS is guaranteed both on the network side which guarantee the low delay of light load and high throughput of heavy load based on path isolation and dynamic path allocation, and host side which proposes a new protocol stack architecture which supports multiple transport protocols and is able to flexibly configure and reconfigure the transmission protocol based on the QoS requirements of the application. In the public cloud to provide tenants-level QoS guarantee, the design is based on the multi-path bandwidth guarantee method and multi-token bucket technology to carry out the multi-path of the speed limit and the dynamic rate adjustment. The research product will lay a necessary theoretical basis for the design of network cloud data center. The optimized cloud data center network will better satisfy the QoS requirements of network applications and cloud tenants, optimize the user experience, improve the utilization rate of the data center network, as well as reduce the cost of cloud computing platform operation.