软件定义网络资源分配与性能优化验证问题研究

软件定义网络（SDN）的优化设计问题是关注焦点，不仅需要有服务质量和性能的考量，更要对消耗和收益进行评估与预测，在SDN的设计中，验证性能边界和局部失效时的自适应响应和性能保证也是瓶颈。本项目从基本理论和实现方法两个角度研究SDN资源优化与性能验证的四个科学问题：研究SDN控制器放置优化问题，提出基于差异化网络流量模式和SDN控制器处理能力的性能模型，减少传播延迟和节点处理延迟；研究SDN网络成本和收益问题，提出性能-消耗-收益三元组博弈模型，评估和预测控制器开启、关闭、扩容等策略下整个网络域中的服务能力和运营商收益的变化；研究SDN网络性能鲁棒验证问题，建立软件定义网络验证框架，提出网络设计性能的验证指标，定量刻画网络情况不确定性；研究SDN网络局部失效响应和多流量模式需求自适应问题，确保网络出现性能需求改变时快速应对，优化服务质量和维持网络性能，提供有效的理论和实际应用方法。

The emergence and spread of the Software Defined Networking (SDN) has brought us possibilities to redesign the network for the provisioning of variety service demands. However, the optimal design of controller placement according to demands is still an open issue. The service providers seek for not only the service quality and performance, but also the cost and benefit. The evaluation and prediction of performance-cost-benefit triad is of concern. Unfortunately, it hardly finds existing works in the open literature which demonstrate the theory or tool of measuring the triad. In addition, another key challenge confronting wide-area SDN architects is the validation of SDN designs providing desired performance in the face of diversified traffic patterns, demands and failures. Validation is challenging because of the exponential, non-enumerable, and multiple scenarios that must be considered. Exiting works measure overly conservative bounds on network performance since to remain tractable. However, the lack of considering flexible routing strategies may not adapt to failures and changing traffic demands. To the best of our knowledge, frameworks which characterize the robust validation of SDN and models that characterize adaption strategies in accordance with failures and traffic patterns are open issues. To this end, this work aims at addressing the following issues: 1 establishing a performance model based on the ingress traffic and SDN controller service capacity, and hence optimizing the controller placement according to the performance metrics, i.e., latency; 2 characterizing the performance-cost-benefit triad model which is employed to evaluate and predict the performance and benefit gained under various network performance; 3 providing an optimization framework to derive the network performance across scenarios, including the capture of performance metrics of network design, the characterization of uncertainty in network conditions, and consequently modelling the adaption strategies respond to the failures and variety traffic demands; 4 emulation on a SDN testbed, including topology generating tool on Mininet, traffic generator, and an actual traffic matrix snapshot in accordance with the validation framework in order to generate set failure scenarios.