A novel approach for the performance bound of QoS-aware data networks under greedy CAC policy

  • Authors:

  • Bahador Bakhshi;Siavash Khorsandi

  • Affiliations:

  • -;-

  • Venue:
  • Performance Evaluation
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

The performance of QoS-aware data networks is measured in terms of acceptance probability of traffic demands requiring an amount of end-to-end bandwidth, which is determined by the Call Admission Control (CAC) policy and the QoS routing algorithm. In this paper, we characterize the theoretical upper bound on the network performance under the greedy CAC policy. Whereas similar problems have been studied in teletraffic networks, the existing approaches consider a specific routing algorithm and obtain its performance. They cannot be used to find performance bound in data networks using complex dynamic QoS routing algorithms. We obtain the bound through modeling the network as a multidimensional Markov chain, which does not depend on a particular routing algorithm. Since this model is computational intensive, the greediest online routing algorithm is developed to estimate the performance bound with lower computational complexity through simulation. In single rate traffic condition, we show that resources allocated by this algorithm for each traffic class is approximately weighted max-min fair. Using this observation, we develop a novel technique to estimate the bound in which an individual loss system is considered per traffic class. Simulation results show that the technique provides an accurate estimation of the bound in a wide range of traffic and network parameter settings.