NETS-NBD: Architectures for Fast Reconfigurable WDM-based Networks

NBD提案0626781随着数据流量的持续增长，网络在传输和处理需求方面都将变得紧张。波分复用（WDM）已经成为一种占主导地位的传输技术;通过在多个波长上实现同时传输，WDM系统实现了超过每秒兆比特的传输容量。然而，在网络节点处以电子方式处理所有这些流量呈现出显著的瓶颈。该项目的目标是通过开发一种高效的网络架构，结合动态网络控制算法，使网络能够有效地响应流量和网络条件的变化，从而缓解电子瓶颈。 该架构采用了电子层路由和WDM层拓扑重构的组合。本计画应用动态系统控制的技术，开发出可最大化网路吞吐量与改善延迟性能的联合重新配置与路由演算法，并借由本计画所开发的架构与演算法，以降低成本的方式设计出可扩充的网路。 它们还使网络能够自主地自我配置，以响应流量条件的变化，从而提高性能。通过这个项目所做的工作是导致新的可重构WDM网络的性能限制和实现这些限制的实用算法的基本理解。拟议活动的更广泛影响包括通过期刊和会议出版物以及网络传播研究成果;将通过这项工作开发的新理论和算法纳入麻省理工学院的研究生课程;以及培训研究生和本科生。

NBD proposal 0626781With the continuing growth in data traffic, networks will be strained in terms of both transport and processing requirements. Wavelength Division Multiplexing (WDM) has emerged as a dominant transport technology; by enabling simultaneous transmission on multiple wavelengths, WDM systems achieve transmission capacities in excess of a Terabit per second. However, electronically processing all of this traffic at network nodes presents a significant bottleneck. The goal of this project is to alleviate the electronic bottleneck by developing an efficient network architecture in combination with dynamic network control algorithms that allow the network to effectively respond to changes in traffic and network conditions. The architecture employs a combination of electronic layer routing and WDM layer topology reconfiguration. Applying techniques from control of dynamical systems, this project develops joint reconfiguration and routing algorithms that maximize network throughput and improve delay performance.The architectures and algorithms developed through this project enable the design of scalable networks at reduced cost. They also enable networks to autonomously self-configure in response to changes in traffic conditions to improve performance. The work done through this project is leading to new fundamental understanding of the performance limits of reconfigurable WDM networks and practical algorithms for achieving these limits. The broader impacts of the proposed activity include the dissemination of the research results through journal and conference publications and the web; incorporation of new theory and algorithms developed through this work into a graduate classes at MIT; and training of graduate and undergraduate students.