面向软件定义核心网的OpenFlow分组转发优先制排队模型研究

软件定义网络（Software-Defined Networking，SDN）作为一种数据转发与控制逻辑相解耦、并开放底层编程接口的创新网络架构，为降低核心网的部署运营成本、提升应用业务性能提供了全新的解决思路.然而，在SDN架构下，逻辑上集中的控制平面容易出现性能瓶颈，进而加大分组转发时延，因此有必要理解其分组转发性能特性.为此，本文首先介绍了软件定义核心网的典型部署场景，分析了控制平面的Packet-in消息到达过程和数据平面的分组到达过程，进而应用M/M/n/m和M/M/1/m排队模型分别刻画控制器集群的Packet-in消息处理过程和OpenFlow交换机的分组处理过程.在此基础上，建立OpenFlow分组转发优先制排队模型，进而推导出不同优先级的分组转发时延及其累积分布函数CDF.最后，借助控制器性能测量工具OFsuite_Performance进行实验评估，结果表明：与现有模型相比，本文所提的M/M/n/m模型更能准确估计控制器集群的实际性能.同时，采用数值分析的方法对比了多种情况下不同优先级的分组转发时延及CDF曲线，为软件定义核心网的实际应用部署提供有效参考.

A Prioritized Queueing Model of OpenFlow Packet Forwarding in Software-Defined Core Networks

As an innovative network architecture decoupling data forwarding and control logic,and opening underlying programming interfaces,software-defined networking (SDN) provides a novel solution to reduce deployment and operation costs and improve business application performance in core networks.However,logically centralized control plane under the SDN architecture is prone to performance bottlenecks,and increases packet forwarding delay.Thus it is necessary to understand the characteristics of its packet forwarding performance.To this end,we first introduce typical deployment scenarios of software-defined core networks,and analyze Packet-in message arrival process in its control plane and packet arrival process in its data plane.Then the M/M/n/m and M/M/1/m queueing models are respectively applied to depict Packet-in message processing process of its controller clusters and packet processing process of its OpenFlow switches.On this basis,we establish a prioritized queueing model of OpenFlow packet forwarding,and derive packet forwarding delays of different priorities and its cumulative distribution function.Finally,experimental evaluation in virtue of the controller performance measurement tool OFsuite_Performance shows that our proposed M/M/n/m model can accurately estimate actual performance of controller clusters compared with existing models.Meanwhile,we contrast packet forwarding delays of different priorities in various cases and their CDF curves by numerical analysis,which provides effective references for practical deployments of software-defined core networks.