TCP flow classification and bandwidth aggregation in optically interconnected data center networks

Optical functionality is being used to realize new data center architectures that minimize electronic switching overheads, pushing the processing to the edge of the network. A challenge in optically interconnected data center networks is to identify the large, bandwidth hungry flows (i.e., elephants) and efficiently establish the optical circuits. Moreover, the amount of optical resources to be provisioned during the network planning phase is a critical design problem. Flow classification accuracy affects the efficiency of optical circuits. Optical channel bandwidth, on the other hand, directly relates to the additive-increase, multiplicative-decrease congestion control mechanism of the transmission control protocol and affects the effective bandwidth allocated to elephant flows. In this paper, we simultaneously investigate the impact of two important mechanisms on data center network performance: traffic flow classification accuracy and optical bandwidth aggregation (i.e., the consolidation of several low-capacity channels into a single high-capacity one by employing advanced modulation formats for short-reach communications). We develop a discrete-event simulator for a hybrid data center network, enabling the tuning of flow classification parameters. Our simulations indicate that data center performance is highly sensitive to the aggregation level.We could observe up to a 74.5% improvement in network throughput only due to consolidating the optical channel bandwidth. We further noticed that the role of flow classification becomes more pronounced with higher bandwidth per wavelength as well as with more hot-spot traffic. Compared to a random classification benchmark, adaptive flow classification could lead to throughput improvements as large as 54.7%.

[1]  Sebastian Zander,et al.  Automated traffic classification and application identification using machine learning , 2005, The IEEE Conference on Local Computer Networks 30th Anniversary (LCN'05)l.

[2]  Dayou Qian,et al.  Design and Evaluation of a Flexible-Bandwidth OFDM-Based Intra-Data Center Interconnect , 2013, IEEE Journal of Selected Topics in Quantum Electronics.

[3]  Hari Balakrishnan,et al.  An experimental study of the learnability of congestion control , 2014, SIGCOMM.

[4]  Ioannis Tomkos,et al.  Optical interconnection networks in data centers: recent trends and future challenges , 2013, IEEE Communications Magazine.

[5]  Ioannis Tomkos,et al.  A Survey on Optical Interconnects for Data Centers , 2012, IEEE Communications Surveys & Tutorials.

[6]  Alex C. Snoeren,et al.  Inside the Social Network's (Datacenter) Network , 2015, Comput. Commun. Rev..

[7]  Vern Paxson,et al.  TCP Congestion Control , 1999, RFC.

[8]  Peter A. Andrekson,et al.  70 Gbps 4-PAM and 56 Gbps 8-PAM Using an 850 nm VCSEL , 2014, Journal of Lightwave Technology.

[9]  Yan Yan,et al.  Griffin: Programmable Optical DataCenter With SDN Enabled Function Planning and Virtualisation , 2015, Journal of Lightwave Technology.

[10]  Ting Wang,et al.  SDN-controlled routing of elephants and mice over a hybrid optical/electrical DCN testbed , 2015, 2015 Optical Fiber Communications Conference and Exhibition (OFC).

[11]  Gil Zussman,et al.  Optical multicast system for data center networks. , 2015, Optics express.

[12]  David A. Maltz,et al.  Network traffic characteristics of data centers in the wild , 2010, IMC '10.

[13]  Madeleine Glick,et al.  Your Data Center Is a Router: The Case for Reconfigurable Optical Circuit Switched Paths , 2009, HotNets.

[14]  Yoshua Bengio,et al.  Understanding the difficulty of training deep feedforward neural networks , 2010, AISTATS.

[15]  Geoffrey E. Hinton,et al.  Visualizing Data using t-SNE , 2008 .

[16]  Amin Vahdat,et al.  Hedera: Dynamic Flow Scheduling for Data Center Networks , 2010, NSDI.

[17]  Jianbo Gao,et al.  TCP AIMD dynamics over Internet connections , 2005, IEEE Communications Letters.

[18]  Roberto Proietti,et al.  Experimental demonstration of dynamic flexible bandwidth optical data center network with all-to-all interconnectivity , 2014, 2014 The European Conference on Optical Communication (ECOC).

[19]  Grenville J. Armitage,et al.  A survey of techniques for internet traffic classification using machine learning , 2008, IEEE Communications Surveys & Tutorials.

[20]  Amin Vahdat,et al.  Helios: a hybrid electrical/optical switch architecture for modular data centers , 2010, SIGCOMM '10.

[21]  Sergey Ioffe,et al.  Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , 2015, ICML.

[22]  Konstantina Papagiannaki,et al.  c-Through: part-time optics in data centers , 2010, SIGCOMM 2010.

[23]  He Liu,et al.  Circuit Switching Under the Radar with REACToR , 2014, NSDI.

[24]  Srinivasan Seshan,et al.  Scheduling techniques for hybrid circuit/packet networks , 2015, CoNEXT.

[25]  Amin Vahdat,et al.  Integrating microsecond circuit switching into the data center , 2013, SIGCOMM.

[26]  Albert G. Greenberg,et al.  The nature of data center traffic: measurements & analysis , 2009, IMC '09.

[27]  Madeleine Glick,et al.  Machine learning based adaptive flow classification for optically interconnected data centers , 2016, 2016 18th International Conference on Transparent Optical Networks (ICTON).

[28]  Praveen Yalagandula,et al.  Mahout: Low-overhead datacenter traffic management using end-host-based elephant detection , 2011, 2011 Proceedings IEEE INFOCOM.

[29]  Sajad Shirali-Shahreza,et al.  Empowering Software Defined Network controller with packet-level information , 2013, 2013 IEEE International Conference on Communications Workshops (ICC).

[30]  Amin Vahdat,et al.  Switching the optical divide: fundamental challenges for hybrid electrical/optical datacenter networks , 2011, SoCC.

[31]  Qin Li,et al.  Extract minimum positive and maximum negative features for imbalanced binary classification , 2012, Pattern Recognit..