Statistical Convergence Analysis of Routing Algorithms

Whenever new and better routing algorithms are developed a comprehensive convergence analysis can show the real achievements of new algorithms. This paper presents a new approach for a convergence analysis together with a new distance vector routing algorithm, which is no longer affected by the well-known Counting-to-Infinity (CTI) problem. The RMTI algorithm uses event-triggered updates instead of time-periodic updates in order to speed up convergence time and to reduce update traffic. Convergence properties of RMTI are compared with RIPv2 under the impeded condition of provoked CTI-situations caused by link failures in order to show the difference of RMTI to common RIPv2 algorithms. The major focus of this paper is the description of a test environment and the approaches used to measure the convergence time of the routing algorithms. Special effort is directed at minimizing measurement perturbations by separation of measurement and evaluation tasks into online data capturing and offline statistical analysis. The results show that this convergence measurement method is a universally valid method and that RMTI is a newly competitive intra domain routing algorithm which can - in contrast to others - execute filtering policies.