摘要
数据中心负载通信特征多种多样,可重构网络可以实现逻辑连接关系的动态变换以匹配不同的应用通信特征,是提高互连网络资源利用率、降低系统能耗、提高灵活性的有效方法之一。传统的电域互连技术很难实现物理拓扑的动态切换,本文基于新兴的光波长选择开关(WSS)技术提出了一种动态可重构光电混合网络结构,在不改变物理连接关系的条件下通过软件配置WSS实现拓扑连接的动态重构和互连带宽的最优调整;基于Misra&Gries算法和Greedy算法提出了面向可重构光网络的快速配置算法,可在满足光器件物理约束的条件下快速求解出典型拓扑对应的光波长配置参数。本文给出了多种物理拓扑的重构过程,1 024个节点的仿真结果显示,合适的拓扑重构对网络带宽的提升超过60%。
Data Centers run many kinds of applications that exhibit various communication patterns among the servers. For a better performance it would be admirable to choose the appropriate topology under different communication patterns. However data centers usually use a single architecture for various applications. In this paper a WSS-based reconfigurable optical network is proposed which dynamically matches the physical topology to various traffic patterns.It significantly improves the link utilization, power efficiency and flexibility of the interconnection network. Besides, a reconfiguration method combining Misra & Griesedge coloring algorithm and Greedy algorithm is recommended. It can calculate the proper wavelength parameters for each wavelength selective switch(WSS) during the reconfiguration process. Simulation results show that proper reconfiguration can be realized by tuning the control plane of WSS and improve the performance(1 024 nodes) beyond 60% under three typical communication patterns.
引文
[1]Benson T,Akella A,Maltz D A.Network traffic characteristics of data centers in the wild[C].In:Proceedings of the 10th ACM SIGCOMM conference on Internet measurement,Melbourne,Australia,2010.267-280
[2]Benson T,Anand A,Akella A,et al.Understanding data center traffic characteristics[C].In:Proceedings of the ACM Workshop on Research on Enterprise Networking,Barcelona,Spain,2009.65-72
[3]Antoniades N,Ellinas G,Homa J,et al.ROADM architectures and WSS implementation technologies[J].Convergence of Mobile and Stationary Next-Generation Networks,2010:643-674.doi:10.1002/9780470630976.ch20
[4]Lea C T.A scalable AWGR-based optical switch[J].Journal of Lightwave Technology,2015,33(22):4612-4621
[5]Ma X,Kuo G S.Optical switching technology comparison:optical MEMS vs.other technologies[J].IEEECommunications Magazine,2003,41(11):16-23
[6]Azizi S,Safaei F,Hashemi N.On the topological properties of Hyper X[J].The Journal of Supercomputing,2013,66(1):572-593
[7]Liu Y H,Zhu M F,Wang J,et al.Xtorus:an extended Torus topology for on-chip massive data communication[C].In:Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops&Ph D Forum(IPDPSW),Shanghai,China,2012.2061-2068
[8]Bergman K,Wang H.Optical Interconnects for Future Data Center Networks[M].New York:Springer,2013.155-167
[9]Lin Y,Anthur A P,O'Duill S,et al.Fast reconfigurable SOA-based all-optical wavelength conversion of QPSK data employing switching tunable pump lasers[C].In:Optical Fiber Communications Conference and Exhibition,Los Angeles,USA,2017.1-3
[10]Ye X,Yin Y,Yoo S J B,et al.DOS:a scalable optical switch for datacenters[C].In:Proceedings of the 6th ACM/IEEE Symposium on Architectures for Networking and Communications Systems,La Jolla,USA,2010.24
[11]Ju H L,Yusoff Z,Belardi W,et al.A tunable WDMwavelength converter based on cross-phase modulation effects in normal dispersion holey fiber[J].Photonics Technology Letters,2003,15(3):437-439
[12]Kang X,Kao Y H,Yang M,et al.Petabit optical switch for data center networks[J].Dca.fee.unicamp.br,2010.doi:10.1007/978-1-4614-4630-9-8
[13]Liboiron-Ladouceur O,Shacham A,Small B A,et al.The data vortex optical packet switched interconnection network[J].Journal of Lightwave Technology,2008,26(13):1777-1789
[14]Sato K I.Realization and application of large-scale fast optical circuit switch for data center networking[J].Journal of Lightwave Technology,2018,36(7):1411-1419
[15]Wang G,Andersen D G,Kaminsky M,et al.c-Through:part-time optics in data centers[J].ACM SIGCOMMComputer Communication Review,2010,40(4):327-338
[16]Farrington N,Porter G,Radhakrishnan S,et al.Helios:a hybrid electrical/optical switch architecture for modular data centers[J].ACM SIGCOMM Computer Communication Review,2010,40(4):339-350
[17]Vizing V G.On an estimate of the chromatic class of a pgraph[J].Diskret Analiz,1964,3:25-30
[18]Bar-Noy A,Motwani R,Naor J.The greedy algorithm is optimal for on-line edge coloring[J].Information Processing Letters,1992,44(5):251-253
[19]Misra J,Gries D.A constructive proof of Vizing's theorem[J].Information Processing Letters,1992,41(3):131-133
[20]Cao Z,Xu J,Chen M,et al.HPPNet Sim:a parallel simulation of large-scale interconnection networks.In:Proceedings of the 2009 Spring Simulation Multiconference,San Diego,USA,2009.32