Multiserver Loss Systems with Subscribers

  • Authors:
  • Ramandeep S. Randhawa;Sunil Kumar

  • Affiliations:
  • McCombs School of Business, The University of Texas at Austin, Austin, Texas 78712;Graduate School of Business, Stanford University, Stanford, California 94305

  • Venue:
  • Mathematics of Operations Research
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

We study a multiserver loss system with two kinds of customers: subscribers and infrequent users. We model the infrequent users' requests for service by a Poisson process. However, noting that the Poisson process is unable to capture repeated interactions as well as retrials, we propose a Markovian on-off-hold model for the subscribers' requests for service that takes into account retrials by subscribers denied service. We analyze this system in an asymptotic regime where the number of subscribers and the arrival rate of the Poisson stream, and consequently the number of servers, grow without bound. In this regime, we identify and prove convergence to diffusion limits for the system. We also prove the convergence of the invariant distributions to the invariant distribution of the limiting diffusions.