Improving the Cost-Effectiveness of WDM Optical Networks

这份提案是乔治华盛顿大学和威廉与玛丽学院的研究人员联合提交的一份提案。尽管WDM(波分复用)光网络具有更低的误码率和更高的容量，但在市场上尚未与其电子同行竞争。主要原因是硬件成本过高，这使得除了少数用户之外的所有用户都无法获得光网络。本文提出的研究针对其中一些令人望而却步的成本，旨在帮助网络设计者回答一些与尺寸和组件大小相关的问题，以帮助降低硬件成本。建议的研究以全光电路交换多光纤长度路由环为起点，这是一种经过测试的技术，并提出了一个问题：能否用更少的硬件获得高性能？虽然网络的许多方面(从可制造性到更高级别的软件设计)构成了总成本，但本项目的重点是通过限制光交换机的复杂性(交叉点)以及通过时隙更有效地使用可用带宽来降低成本。我们将研究由这些成本考虑因素产生的三种类型的资源分配问题，并设计和评估执行资源分配的算法。第一类问题涉及限制光纤到光纤交换机的灵活性，以努力降低交换机的制造成本。初步结果表明，即使显著降低了交换机的连接性，也可以获得接近最佳的吞吐量。在这种情况下将解决的一些研究问题是：在切换到光纤的限制下，性能如何在数量上发生变化？波长分配与这些限制有根本区别吗？第二类问题涉及从较小的环构建更大的网络。这个问题的研究问题包括：如何以经济高效的方式连接全光多环，特别是通过使用光纤交换限制？对于环间和环内会话，波长应该如何划分(如果有的话)？第三种类型的问题涉及在具有时隙的光网络中有效地调度多速率会话(时分多路复用)。该技术的实际可行性已经得到了论证。这里的重点是资源分配问题，通过更有效地使用带宽和光纤交换限制，降低整体网络成本。最后，将采取综合方法来研究当所有三种类型的设计特征都存在时的网络性能。研究结果将在不久的将来应用，因此最初的方法将采用模拟和建模相结合的方法，而稍后的PI将致力于对问题进行更深层次的理论理解。

This proposal is a joint proposal submitted by investigators at the George Washington University and the College of William and Mary. In spite of lower error rates and far higher capacities, WDM (Wavelength Division Multiplexing) optical networks have yet to compete with their electronic counterparts in the marketplace. The main reason is the prohibitive cost of the hardware, which puts optical networks out of reach for all but a few users.The research proposed here takes aim at some of these prohibitive costs and is intended to help network designers answer some questions related to dimensioning and component-sizing that can help reduce hardware costs.The proposed research takes as a starting point an all-optical circuit-switched multifiber avelength-routing ring, a tested technology, and asks the question ``Can high performance be obtained with less hardware?'' While many aspects of a network (ranging from manufacturability to higher-level software design) make up the overall cost, this project focuses on reducing costs by limiting the complexity (crosspoints) of optical switches and from more effectively using available bandwidth by time-slotting.Three types of resource allocation problems that arise from these cost considerations will be investigated and algorithms that perform resource allocation will be designed and evaluated.The first type of problem involves limiting the flexibility of fiber-to-fiber switches in an effort to reduce the cost of manufacturing switches. Preliminary results indicate that near-optimal throughput can be obtained even with significant reduction in the connectivity of the switch. Some research questions that will be addressed in this case are: how does performance vary quantitatively with limitations in switching to fibers? does wavelength allocation differ fundamentally with these restrictions?The second type of problem involves constructing larger networks from smaller rings. Research questions for this problem include: how should all-optical multiple rings be connected in a cost-effective manner, especially by using fiber-switching limitations? how should wavelengths be partitioned (if at all) for inter-ring and intra-ring sessions?The third type of problem involves efficiently scheduling multi-rate sessions in an optical network with slots (time-division multiplexing). The practical feasibility of this technology has already been demonstrated. The focus here is on resource allocation problems that, by more effectively using bandwidth with fiber-switching limitations, reduce overall network cost.Finally, an integrative approach will be taken to study network performance when all three types of design features are present. The research results are intended to be applicable in the near future, and therefore the initial approach will use a combination of simulation and modeling, whereas later the PIs will aim for a deeper theoretical understanding of the problem.