一种实用的多目标透明WDM光网络设计方法
A pragmatic approach for designing transparent WDM optical networks with multi-objectives
摘要：
面对互联网流量的爆炸性增长，基于WDM技术的光传送网构成了承载多Tb/s的互联网基础设施的核心部分，必须从设计、规划、运营和管理的角度重新考虑，以提高效率。由于光传输技术和光子交换技术的重大进展的融合，透明（全光）架构已经付诸实践，为消除昂贵的光-电-光（O-E-O）接口的过度利用铺平了道路，因此，与不透明架构相比，可以显著节省成本和能耗。传统的透明光网络设计基于单目标优化模型，旨在优化一个单一的性能指标似乎是不够的，以捕捉实际设计的细微差别，而传统的多目标方法往往达到（非）最佳的解决方案。不同于现有的作品，我们提出了一个新的框架，多目标WDM网络设计捕获几个目标，一方面，另一方面，实现最佳的解决方案。此外，我们的建议，利用每个组成目标的特点，奠定了基础，设置权重系数，从而保证了优化的顺序。同样重要的是，我们的建议是务实的意义上，优化模型的复杂性保持不变的单目标模型，而解决方案的质量得到了很大的提高。我们已经广泛地测试了现实的光核心网络拓扑结构，即，COST 239和NSFNET，与各种网络流量条件，事实证明，我们的设计带来了节省波长链路的使用高达约$28\%$在最有利的情况下，而$14\%$预计最不利的情况下。
英文摘要：
In facing with the explosive Internet traffic growth, optical transport networks based on WDM technologies forming the core part of Internet infrastructure carrying multi-Tb/s has to be re-considered from both designing, planning, operation and management perspectives to attain greater efficiency. Thanks to the convergence of significant advances in optical transmission technologies, and photonic switching, transparent (all-optical) architecture has come into practice, paving the way for eliminating the over-utilization of costly optical-electrical-optical (O-E-O) interfaces and hence, yielding remarkable savings of cost and energy consumption compared to opaque architecture. Traditional designs for transparent optical networks based on single-objective optimization model aiming at optimizing solely a single performance metric appears to be insufficient to capture the nuances of practical designs while conventional multi-objective approach tends to reach (non-) optimal solutions. Different from existing works, we present a new framework for multi-objective WDM network designs capturing several goals on one hand and on the other hand, achieving optimal solutions. Moreover, our proposal exploits the characteristics of each constituent objectives to lay the foundation for setting up weight coefficient so that the order of optimization is guaranteed. Equally important, our proposal is pragmatic in the sense that the complexity of the optimization model remains the same as the single-objective model while the quality of solution has been greatly improved. We have extensively tested realistic optical core networks topologies, that is, COST239 and NSFNET, with various network traffic conditions and it turns out that our design brings about a saving of wavelength link usage up to roughly $28\%$ in the most favorable cases while $14\%$ is expected for the least favorable cases.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
