Research Article
A Translucent OBS Node Architecture to Improve Traffic Emission and Loss Probability
@INPROCEEDINGS{10.4108/ICST.BROADNETS2009.7869, author={Thomas Coutelen and Brigitte Jaumard and G\^{e}rard H\^{e}buterne}, title={A Translucent OBS Node Architecture to Improve Traffic Emission and Loss Probability}, proceedings={2nd International ICSTl Workshop on Optical Burst/Packet Switching}, publisher={IEEE}, proceedings_a={WOBS}, year={2009}, month={11}, keywords={Delay Optical buffering Optical burst switching Optical losses Protocols Quality of service Switching circuits Telecommunication switching Telecommunication traffic Traffic control}, doi={10.4108/ICST.BROADNETS2009.7869} }
- Thomas Coutelen
Brigitte Jaumard
Gérard Hébuterne
Year: 2009
A Translucent OBS Node Architecture to Improve Traffic Emission and Loss Probability
WOBS
IEEE
DOI: 10.4108/ICST.BROADNETS2009.7869
Abstract
All-optical circuit switching prevents contention by forbidding multiplexing beyond the wavelength granularity. The drawbacks of such a coarse granularity can be reduced thanks to translucent architectures with MSPPs (MultiService Provisioning Platforms). Another switching paradigm of interest is offered by the Optical Burst Switching (OBS) that manages finer granularity transfers but that encounters contentions, potentially entailing high loss rates. In this paper, within the context of core networks, we describe a translucent OBS architecture that performs intermediate re-aggregation and allows the reduction of the loss rate by interceding at two levels: Firstly, it provides sparse electrical buffering accessibility and, secondly, it uses aggregation grooming in order to reduce the aggregation delay and the traffic contention probability. A careful analysis of the traffic behavior in core networks leads us to propose an accurate traffic model, called LCH+, i.e., an enhanced Lost Call Held (LCH) model. Data plane transparency is then exploited to modulate the traffic at its emission. Extensive experiments show that re-aggregation helps reducing the loss probability thanks to electrical buffers whereas the aggregation grooming mitigates the drawbacks induced by electrical buffering on the delay and on the network cost. In addition, we show that the proposed translucent architecture can be highly valuable in clustered large core networks.