Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic systemCitation formats

  • External authors:
  • Alexander Rast
  • Javier Navaridas Palma
  • X Jin
  • F Galluppi
  • J Miguel-Alonso
  • C Patterson
  • Mikel Lujan

Standard

Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic system. / Rast, Alexander; Navaridas Palma, Javier; Jin, X; Galluppi, F; Plana, Luis A.; Miguel-Alonso, J; Patterson, C; Lujan, Mikel; Furber, Stephen.

In: International Journal of Parallel Programming, Vol. 40, No. 6, 12.2012, p. 553-582.

Research output: Contribution to journalArticlepeer-review

Harvard

Rast, A, Navaridas Palma, J, Jin, X, Galluppi, F, Plana, LA, Miguel-Alonso, J, Patterson, C, Lujan, M & Furber, S 2012, 'Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic system', International Journal of Parallel Programming, vol. 40, no. 6, pp. 553-582. https://doi.org/10.1007/s10766-011-0180-7

APA

Rast, A., Navaridas Palma, J., Jin, X., Galluppi, F., Plana, L. A., Miguel-Alonso, J., Patterson, C., Lujan, M., & Furber, S. (2012). Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic system. International Journal of Parallel Programming, 40(6), 553-582. https://doi.org/10.1007/s10766-011-0180-7

Vancouver

Rast A, Navaridas Palma J, Jin X, Galluppi F, Plana LA, Miguel-Alonso J et al. Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic system. International Journal of Parallel Programming. 2012 Dec;40(6):553-582. https://doi.org/10.1007/s10766-011-0180-7

Author

Rast, Alexander ; Navaridas Palma, Javier ; Jin, X ; Galluppi, F ; Plana, Luis A. ; Miguel-Alonso, J ; Patterson, C ; Lujan, Mikel ; Furber, Stephen. / Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic system. In: International Journal of Parallel Programming. 2012 ; Vol. 40, No. 6. pp. 553-582.

Bibtex

@article{080292ae1ce741e5b57df8cd87d4980d,
title = "Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic system",
abstract = "Neural networks present a fundamentally different model of computation from the conventional sequential digital model, for which conventional hardware is typically poorly matched. However, a combination of model and scalability limitations has meant that neither dedicated neural chips nor FPGA's have offered an entirely satisfactory solution. SpiNNaker introduces a different approach, the {"}neuromimetic{"} architecture, that maintains the neural optimisation of dedicated chips while offering FPGA-like universal configurability. This parallel multiprocessor employs an asynchronous event-driven model that uses interrupt-generating dedicated hardware on the chip to support real-time neural simulation. Nonetheless, event handling, particularly packet servicing, requires careful and innovative design in order to avoid local processor congestion and possible deadlock. We explore the impact that spatial locality, temporal causality and burstiness of traffic have on network performance, using tunable, biologically similar synthetic traffic patterns. Having established the viability of the system for real-time operation, we use two exemplar neural models to illustrate how to implement efficient event-handling service routines that mitigate the problem of burstiness in the traffic. Extending work published in ACM Computing Frontiers 2010 with on-chip testing, simulation results indicate the viability of SpiNNaker for large-scale neural modelling, while emphasizing the need for effective burst management and network mapping. Ultimately, the goal is the creation of a library-based development system that can translate a high-level neural model from any description environment into an efficient SpiNNaker instantiation. The complete system represents a general-purpose platform that can generate an arbitrary neural network and run it with hardware speed and scale. {\textcopyright} The Author(s) 2011.",
keywords = "Asynchronous, Burst, Characterisation, Event-driven, Interconnection, Multiprocessor, Network, Neural, Real-time, Traffic, Universal",
author = "Alexander Rast and {Navaridas Palma}, Javier and X Jin and F Galluppi and Plana, {Luis A.} and J Miguel-Alonso and C Patterson and Mikel Lujan and Stephen Furber",
note = "cited By 2",
year = "2012",
month = dec,
doi = "10.1007/s10766-011-0180-7",
language = "English",
volume = "40",
pages = "553--582",
journal = "International Journal of Parallel Programming",
issn = "0885-7458",
publisher = "Springer Nature",
number = "6",

}

RIS

TY - JOUR

T1 - Managing burstiness and scalability in event-driven models on the spinnaker neuromimetic system

AU - Rast, Alexander

AU - Navaridas Palma, Javier

AU - Jin, X

AU - Galluppi, F

AU - Plana, Luis A.

AU - Miguel-Alonso, J

AU - Patterson, C

AU - Lujan, Mikel

AU - Furber, Stephen

N1 - cited By 2

PY - 2012/12

Y1 - 2012/12

N2 - Neural networks present a fundamentally different model of computation from the conventional sequential digital model, for which conventional hardware is typically poorly matched. However, a combination of model and scalability limitations has meant that neither dedicated neural chips nor FPGA's have offered an entirely satisfactory solution. SpiNNaker introduces a different approach, the "neuromimetic" architecture, that maintains the neural optimisation of dedicated chips while offering FPGA-like universal configurability. This parallel multiprocessor employs an asynchronous event-driven model that uses interrupt-generating dedicated hardware on the chip to support real-time neural simulation. Nonetheless, event handling, particularly packet servicing, requires careful and innovative design in order to avoid local processor congestion and possible deadlock. We explore the impact that spatial locality, temporal causality and burstiness of traffic have on network performance, using tunable, biologically similar synthetic traffic patterns. Having established the viability of the system for real-time operation, we use two exemplar neural models to illustrate how to implement efficient event-handling service routines that mitigate the problem of burstiness in the traffic. Extending work published in ACM Computing Frontiers 2010 with on-chip testing, simulation results indicate the viability of SpiNNaker for large-scale neural modelling, while emphasizing the need for effective burst management and network mapping. Ultimately, the goal is the creation of a library-based development system that can translate a high-level neural model from any description environment into an efficient SpiNNaker instantiation. The complete system represents a general-purpose platform that can generate an arbitrary neural network and run it with hardware speed and scale. © The Author(s) 2011.

AB - Neural networks present a fundamentally different model of computation from the conventional sequential digital model, for which conventional hardware is typically poorly matched. However, a combination of model and scalability limitations has meant that neither dedicated neural chips nor FPGA's have offered an entirely satisfactory solution. SpiNNaker introduces a different approach, the "neuromimetic" architecture, that maintains the neural optimisation of dedicated chips while offering FPGA-like universal configurability. This parallel multiprocessor employs an asynchronous event-driven model that uses interrupt-generating dedicated hardware on the chip to support real-time neural simulation. Nonetheless, event handling, particularly packet servicing, requires careful and innovative design in order to avoid local processor congestion and possible deadlock. We explore the impact that spatial locality, temporal causality and burstiness of traffic have on network performance, using tunable, biologically similar synthetic traffic patterns. Having established the viability of the system for real-time operation, we use two exemplar neural models to illustrate how to implement efficient event-handling service routines that mitigate the problem of burstiness in the traffic. Extending work published in ACM Computing Frontiers 2010 with on-chip testing, simulation results indicate the viability of SpiNNaker for large-scale neural modelling, while emphasizing the need for effective burst management and network mapping. Ultimately, the goal is the creation of a library-based development system that can translate a high-level neural model from any description environment into an efficient SpiNNaker instantiation. The complete system represents a general-purpose platform that can generate an arbitrary neural network and run it with hardware speed and scale. © The Author(s) 2011.

KW - Asynchronous

KW - Burst

KW - Characterisation

KW - Event-driven

KW - Interconnection

KW - Multiprocessor

KW - Network

KW - Neural

KW - Real-time

KW - Traffic

KW - Universal

U2 - 10.1007/s10766-011-0180-7

DO - 10.1007/s10766-011-0180-7

M3 - Article

VL - 40

SP - 553

EP - 582

JO - International Journal of Parallel Programming

JF - International Journal of Parallel Programming

SN - 0885-7458

IS - 6

ER -