Performance analysis of alternate routing with trunk reservation in multirate switched networks

  • Authors:
  • Joonhwan Kim;Donghwan Lee;Saewoong Bahk

  • Affiliations:
  • School of Electrical Engineering, Seoul National University, Seoul, South Korea;School of Electrical Engineering, Seoul National University, Seoul, South Korea;School of Electrical Engineering, Seoul National University, Seoul, South Korea

  • Venue:
  • Computer Communications
  • Year:
  • 2002

Quantified Score

Hi-index 0.24

Visualization

Abstract

The traditional approach to implementing call routing in PSTN networks uses trunk reservation to achieve high throughput by avoiding possible routing oscillations in dynamic alternate routing. The purpose of trunk reservation in PSTN networks that support only a single class of traffic, e.g. voice is to limit the use of alternate paths for call routing when the network is heavily loaded. In this paper, we model the system of dynamic alternate routing with trunk reservation for multirate switched networks where the Markov chain is not time reversible, and analyze it by using an iteration method. Then as the calculation to find the optimal reservation size is intractable, we slightly modify the conventional Kaufman's recursive method that has been applied for the reversible Markov chain system. The complexity of the modified approximation is the same as that of the Kaufman's method. For a network of which link capacity is 50 channels, the approximation shows quite similar results to the analysis. For a network of which link capacity is 750 channels, which state size is not workable, we compare our approximation with simulation results. If we count some computation errors caused by the reduced load effect in the approximation, we can conclude that our approximation is quite accurate. As alternate routing with trunk reservation is expected to run on the top of multirate switched networks, our proposed approach can be used to obtain the near optimal trunk reservation size in real time, which results in maximal throughput.