High-performance networks: traffic, protocols, and control algorithms

我的研究方案涉及高性能分组网络中的流量、协议和控制算法的分析和模拟。分组数据网络(如Internet)是传输语音、数据和视频应用的基础设施。它们的性能取决于网络元件(路由器)中采用的网络流量、协议和调度机制的复杂交互。评估这一性能是一项困难的任务，对服务提供商和网络用户都很重要。我的研究目标是首先对从部署的分组网络收集和测量的流量进行特征、建模和分析。特别感兴趣的是应用统计、数据挖掘和聚类方法来搜索流量不变量，以及理解代表承载这种流量的分组网络的复杂系统的潜在动态行为。建议的研究计划包括三个子领域：从部署的网络收集的流量数据的分析和表征，网络协议和算法的模拟和改进以及网络性能的评估，以及控制理论在复杂动态网络行为分析中的应用。收集的流量数据将使我们能够检查流量数据收集、用户行为和网络操作的各个方面，并评估部署的路由协议的性能。将开发模拟工具和模型，以评估和改进各种有线和无线网络协议和算法的性能。开发的网络协议和机制的数学模型将被用来更好地理解在TCP/IP网络模型中观察到的复杂行为，如混沌和分叉现象。

My research proposal deals with analysis and simulation of traffic, protocols, and control algorithms in high-performance packet networks.Packet data networks, such as the Internet, are the infrastructure for delivering voice, data, and video applications. Their performance depends on a complex interaction of network traffic, protocols, and scheduling mechanisms employed in network elements (routers). Evaluating this performance is a difficult task that is of importance both to service providers and network users. The objective of my research is to first characterize, model, and analyze traffic that is collected and measured from deployed packet networks. Of particular interest is application of statistical, data mining, and clustering approaches in the search for traffic invariants as well as the understanding of underlying dynamical behavior of the complex system representing a packet network carrying such traffic.The proposed research program covers three subareas: analysis and characterization of traffic data collected from deployed networks, simulation and improvement of network protocols and algorithms and evaluation of network performance, and application of control theory towards the analysis of complex dynamic network behavior. Collected traffic data will enable us to examine various aspects of traffic data collection, user behavior, and network operation and to evaluate performance of deployed routing protocols. Simulation tools and models will be developed to evaluate and improve performance of various wired and wireless network protocols and algorithms. Developed mathematical models of network protocols and mechanisms will be used to better understand complex behavior, such as chaotic and bifurcation phenomena, that has been observed in models of TCP/IP networks.