Abstract
We consider low-complexity, all-optical multi-stage networks with distributed arbitration achieved through minimal per-node buffering. To enhance the saturation throughput of such networks while maintaining low latency at low loads, we examine a novel combination of deterministic (prescheduled) and speculative (eager) packet injections. Prescheduled injections are performed in a time-division-multiplexing (TDM) manner, whereas eager injections follow a packet multiplexing paradigm. Prescheduled injections, on the one hand, aim at reducing contention in the fabric, and sustain network throughput when the load is high. Eager injections, on the other hand, ignore the TDM schedule, thus allowing low-latency communication when contention is low. The rules that govern the interaction between prescheduled and eager packets allow eager packets to be dropped intentionally when they block the progress of prescheduled ones. At the same time, a highly efficient end-to-end reliable delivery scheme, implemented at the network interfaces, deals with packet losses in the optical domain and also recovers the dropped eager packets. Computer simulations demonstrate that our approach can render low-complexity networks, which are amenable to an all-optical implementation, attractive for use in computer interconnects.
∗ Part of this chapter was published previously in Proc. IEEE 12th International Conference on High Performance Switching and Routing (HPSR), Dallas, TX (2010).IBM, the IBM logo, and ibm.com are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. Other product and service names might be trademarks of IBM or other companies.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
As NACKs should be generated within the optical fabric, their implementation would not be feasible in the proposed architecure.
- 2.
Simulation models and parameters are described in Sect. 6.4.
- 3.
We restrict this work to uniform traffic, as Omega networks are inherently incapable of routing all possible non-uniform traffic patterns because of internal conflicts. In principle, other matching sequences are possible, and the applied matching sequence could be changed dynamically.
- 4.
This is equal to the next-awaited packet ID “minus 1” mod 2W.
- 5.
This is the injection policy that we use in our computer simulations. A slightly modified version of it is presented in Sect. 6.4.4.
- 6.
Not all unacked packets are included in B, but only those that the adapter has to retransmit.
- 7.
ACKs have the lowest priority, because any injected packet from source s to destination d piggybacks ACK information for the reverse flow d → s.
- 8.
We give precedence to the prescheduled flow, although the injected packet will be eager, to reduce the conflicts with prescheduled packets coming from other adapters.
References
Beldianu S, Rojas-Cessa R, Oki E, Ziavra S (2009) Re-configurable parallel match evaluators applied to scheduling schemes for input-queued packet switches. In: Proceedings of IEEE ICCCN, San Francisco, CA, USA
Blumenthal DJ et al (2011) Integrated photonics for low-power packet networking. IEEE J Sel Top Quant Electron 17(2):458–471
Burmeister EF, Blumenthal DJ, Bowers JE (2008) A comparison of optical buffering technologies. Optical Switching and Networking 5(1):10–18
Chang CS, Lee DS, Jou YS (2002) Load-balanced Birkhoff-von Neumann switches, part I: one-stage buffering. Comp Comm 25(6):611–622
Chao HJ, Jing Z, Liew SY (2003) Matching algorithms for three-stage bufferless clos network switches. IEEE Comm Mag 41:46–54 (2003)
Chrysos N, Katevenis M (2006) Scheduling in non-blocking, buffered, three-stage switching fabrics. In: Proceedings of IEEE INFOCOM. Barcelona, Spain
Dias D, Jump JR (1981) Analysis and simulation of buffered delta networks. IEEE Trans Comput C-30(4):273–282
Germann R, Salemink HWM, Beyeler R, Bona GL, Horst F, Massarek I, Offrein BJ (2000) Silicon oxynitride layers for optical waveguide applications. J Electrochem Soc 147(6):2237–2241
Goke LR, Lipovski GJ (1973) Banyan networks for partitioning multiprocessor systems. In: Proceedings of ACM ISCA, New York, NY, USA, pp 21–28
Hui JH (1990) Switching and traffic theory for integrated broadband networks. Kluwer, Dordrecht
Iliadis I, Minkenberg C (2008) Performance of a speculative transmission scheme for arbitration latency reduction. IEEE/ACM Trans Comp 16(1):182–195
Iliadis I, Chrysos N, Minkenberg C (2007) Performance evaluation of the data vortex photonic switch. IEEE J Sel Areas Comm 25(S-6):20–35
Keslassy I, Chuang ST, Yu K, Miller D, Horowitz M, Solgaard O, McKeown N (2003) Scaling internet routers using optics. In: Proceedings of ACM SIGCOMM. ACM, Karlsruhe, pp 189–200
Liu J, Hung CK, Hamdi M, Tsui CY (2002) Stable round-robin scheduling algorithms for high-performance input queued switches. In: Proceedings of IEEE hot-interconnects (HOTI 2002), San Francisco, CA
Luijten, RP, Minkenberg C, Hemenway BR, Sauer M, Grzybowski R (2005) Viable opto-electronic HPC interconnect fabrics. In: Proceedings of supercomputing (SC). IEEE Computer Society, Washington, DC (2005)
Murdocca M (1989) Optical design of a digital switch. Appl Opt 28(13):2505–2517
Papadimitriou GI, Papazoglou C, Pomportsis AS (2003) Optical switching: switch fabrics, techniques, and architectures. J Lightwave Technol 21(2):384–405
Petracca M, Lee BG, Bergman K, Carloni LP (2009) Photonic nocs: system-level design exploration. IEEE Micro 29(4), 74–85 (2009)
Pun K, Hamdi M (2002) Distro: A distributed static round-robin scheduling algorithm for bufferless clos-network switches. In: Proceedings of IEEE GLOBECOM, Taipei, Taiwan, pp 2298–2302
Saha A, Wagh M (1990) Performance analysis of banyan networks based on buffers of various sizes. In: IEEE INFOCOM, San Francisco, CA, pp. 157–164
Scicchitano A, Bianco A, Giaccone P, Leonardi E, Schiattarella E (2007) Distributed scheduling in input queued switches. In: Proceedings of IEEE ICC, Glasgow, UK
Shacham A, Small BA, Liboiron-Ladouceur O, Bergman K (2005) A fully implemented 12 x 12 data vortex optical packet switching interconnection network. J Lightwave Technol 23(10):3066
Takagi H (1993) Queueing analysis. In: Discrete-time systems: a foundation of performance evaluation. Elsevier, Amsterdam
Tanenbaum AS (2002) Computer networks, 4th edn. Prentice Hall, NJ
Acknowledgements
This research was supported by the European Union FP7-ICT project HISTORIC (“Heterogeneous InP on Silicon Technology for Optical Routing and logIC”) under contract no. 223876. The authors would like to thank Anne-Marie Cromack and Charlotte Bolliger for their help in preparing this manuscript.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2013 Springer Science+Business Media New York
About this chapter
Cite this chapter
Chrysos, N., Hofrichter, J., Horst, F., Offrein, B., Minkenberg, C. (2013). All-Optical Networks: A System’s Perspective. In: Kachris, C., Bergman, K., Tomkos, I. (eds) Optical Interconnects for Future Data Center Networks. Optical Networks. Springer, New York, NY. https://doi.org/10.1007/978-1-4614-4630-9_6
Download citation
DOI: https://doi.org/10.1007/978-1-4614-4630-9_6
Published:
Publisher Name: Springer, New York, NY
Print ISBN: 978-1-4614-4629-3
Online ISBN: 978-1-4614-4630-9
eBook Packages: EngineeringEngineering (R0)