6th International ICST Conference on Broadband Communications, Networks, and Systems

Research Article

Complexity Analysis and Performance Evaluation of a Two-Step Scheduler for Modular Optical Packet Switches

Download490 downloads
  • @INPROCEEDINGS{10.4108/ICST.BROADNETS2009.7268,
        author={P. G. Raponi and N.  Andriolli and P. Castoldi and A. Bianchi},
        title={Complexity Analysis and Performance Evaluation of a Two-Step Scheduler for Modular Optical Packet Switches},
        proceedings={6th International ICST Conference on Broadband Communications, Networks, and Systems},
        publisher={IEEE},
        proceedings_a={BROADNETS},
        year={2009},
        month={11},
        keywords={Delay Network servers Optical packet switching Optical receivers Optical switches Optical transmitters Performance analysis Processor scheduling Wavelength routing Web server},
        doi={10.4108/ICST.BROADNETS2009.7268}
    }
    
  • P. G. Raponi
    N. Andriolli
    P. Castoldi
    A. Bianchi
    Year: 2009
    Complexity Analysis and Performance Evaluation of a Two-Step Scheduler for Modular Optical Packet Switches
    BROADNETS
    IEEE
    DOI: 10.4108/ICST.BROADNETS2009.7268
P. G. Raponi1,*, N. Andriolli1, P. Castoldi1, A. Bianchi2
  • 1: Scuola Superiore Sant’Anna, Pisa, Italy
  • 2: Ericsson, Pisa, Italy
*Contact email: raponi@sssup.it

Abstract

In this paper we evaluate the complexity and the performance of two different schedulers for high-speed optical switches with a large number of ports. We compare the classical approach with a two-step scheduling framework recently proposed for multi-card optical switches in which firstly transmitters on each card are assigned a different wavelength, then for every wavelength a maximal matching is found among all cards. We demonstrate that the two-step scheduler is characterized by a lower complexity and it is thus more scalable than the classical approach. Then we analyze the performance of the proposed framework under bursty traffic and the benefits introduced by the use of transfer speedup, a technique which allows to send more than one packet in the same time slot, while issuing a single scheduling decision. Results obtained by simulations on a wide range of cards and wavelengths, show that the two-step scheduling framework achieves a performance comparable to the classical approach at high loads. Moreover the introduction of a small transfer speedup is shown to significantly reduce the average cell latency.