基于非等粒度光树的数据中心光网络流聚合技术研究

Optical datacenter network has become the basic physical facility of the Internet, Cloud Computing, and Big Data. It has obtained extensive research by academia and Industry. A light-tree is usually used to serve a multicast demand when multicast services are gaining popularity and momentum. Since the spectrum bandwidth on each link of an conventional equal granularity light-tree (EGLT) are equal, so that the EGLT based multicast flow aggregation scheme has the disadvantages of spectrum resource waste, limited aggregation capability and adaptive ability, lacked survivability against disasters, especially in a network scenario with a lot of asymmetric or unauthorized users, local limited spectrum resources, and distributed service provisioning..To address this problem, we design a novel non-equal granularity light-tree (NeGLt) which supports the user-desired spectrum allocation on each link. NeGLt ensures that each user can only obtain the desired multicast services. It can eliminate the spectrum waste and improve the aggregation capability and adaptive ability of multicast flow aggregation scheme. This project studies the non-equal granularity light-tree based multicast flow aggregation (NeGLt-MFA) scheme and designs a novel node architecture to implement the NeGLt in optical datacenter networks. The NeGLt-MFA scheme provides a spectrum-efficient multicast service transmission approach for a large number of multicast demands, especially in a network scenario with a lot of asymmetric or unauthorized users, local limited spectrum resources, and distributed service provisioning. The proposed NeGLt-MFA scheme can improve the resource efficiency and service transmission capacity of optical datacenter networks.

数据中心光网络已成为互联网、云计算，以及大数据等产业重要的基础物理设施，受到学术界和产业界的广泛研究。针对大量涌现的多播业务，利用光树进行多播业务的承载逐渐被应用起来。由于传统等粒度光树的每条链路传输相等频谱宽度的光信号，基于等粒度光树的流聚合技术在面对大量非对称用户、指定授权用户、局部有限频谱资源、分布式业务提供时，呈现频谱资源浪费、聚合能力受限、自适应能力弱、抗毁能力不足等缺陷。.为应对上述问题，本项目提出了一种新型的光传输结构：非等粒度光树。非等粒度光树保证每个用户只会收到其所请求的业务，能够消除频谱浪费，提高流聚合技术的聚合能力和自适应能力。本项目研究基于非等粒度光树的流聚合技术，设计支持非等粒度光树的节点结构，进行非等粒度光树的构建，在面临大量非对称用户、指定授权用户、局部有限频谱资源，以及分布式业务提供时提供高效的业务传输手段，提高数据中心光网络的资源效率和业务承载能力。

本课题研究数据中心光网络多播业务模型与内容连通理论、面向非等粒度光树的流聚合技术、非等粒度光树的构建，以及支持非等粒度光树的节点结构，在“业务模型与内容连通理论”方面建立了多播业务聚合的非等粒度光树模型、多播业务承载的分布式子树模型，以及基于内容连通的多播共享保护模型；在“面向非等粒度光树的流聚合技术”方面设计实现了基于非等粒度光树的多播流聚合策略和基于分布式子树的多播聚合策略，最小化频谱和光收发机的消耗；在“非等粒度光树的构建”方面设计了按照最长分支长度的业务编排策略，进行按需的频谱资源分配，将非等粒度光树上因多播聚合带来的频谱冗余将为零；在“支持非等粒度光树的节点结构”方面利用ROADM弹性的频谱选择功能，设计了基于ROADM的支持非等粒度光树的节点结构。本项目取得了一系列具有国际重要影响的研究成果，在IEEE Transactions on Communications 、Optics Express、Optical Fiber Technology等国际权威期刊上累计发表标注的SCI检索论文9篇，EI检索国际会议论文10篇，申请国家发明专利7项，获授权国家发明专利5项，提交IETF标准组织草案3篇，特邀报告1次，出版学术专著1部，获得ICOCN 2018 Best Paper Award（第二作者），2019中国通信学会科学技术奖一等奖（排名第三），获得Optical Switching and Networking 期刊2019 Fabio Neri Best Paper Award（第三作者）。此外，本项目累计培养博士研究生5名、硕士研究生7名，完成了既定的预期目标。