2nd International ICST Workshop on Game Theory in Communication Networks

Research Article

Effects of Parameters on Nash Games with OSNR Target

  • @INPROCEEDINGS{10.4108/ICST.VALUETOOLS2008.4432,
        author={Yan Pan and Tansu Alpcan and Lacra Pavel},
        title={Effects of Parameters on Nash Games with OSNR Target},
        proceedings={2nd International ICST Workshop on Game Theory in Communication Networks},
        publisher={ACM},
        proceedings_a={GAMECOMM},
        year={2010},
        month={5},
        keywords={},
        doi={10.4108/ICST.VALUETOOLS2008.4432}
    }
    
  • Yan Pan
    Tansu Alpcan
    Lacra Pavel
    Year: 2010
    Effects of Parameters on Nash Games with OSNR Target
    GAMECOMM
    ICST
    DOI: 10.4108/ICST.VALUETOOLS2008.4432
Yan Pan1,*, Tansu Alpcan2,*, Lacra Pavel1,*
  • 1: Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 3G4, Canada
  • 2: Deutsche Telekom Labs, Ernst-Reuter-Platz 7, D-10587 Berlin, Germany
*Contact email: yanpan@control.utoronto.ca, tansu.alpcan@telekom.de, pavel@control.toronto.edu

Abstract

This paper studies efficiency in a Nash game with optical signal-to-noise ratio (OSNR) target. Instead of looking from the view point of degree of efficiency ("price of anarchy"), we investigate the effects of parameters in individual cost functions. We show that the aggregate cost function in the game-theoretic formulation is not automatically convex and the optimal solution of the associated constrained optimization problem is not immediate. Then we build the relation between these two formulations by indicating that the individual cost function Ci(ui) in the system optimization formulation has an approximate interpretation with the one Ji(u) in the game-theoretic formulation. We compare simulation results from both a system optimization and a user optimization (game-theoretic) approach for a single optical link. It is well known that the Nash equilibria of a game may not achieve full efficiency. We show the effects of pricing mechanisms on system performance. We also show that OSNR target can be achieved and efficiency can be possibly improved by appropriate selection of parameters.