A workload analysis tool for discrete-time semi-Markovian servers

Abstract

In this paper, we describe the software toolkit Inter-VerdiKom which implements a variety of reliable workload analysis techniques for queues with general or semi-Markovian arrival processes used in stochastic traffic modeling. We compute the workload at a network element in an open queuing network using transient and steady state analysis methods in interval arithmetic. We also present a modeling approach that employs a genetic programming optimization technique, providing an accurate model for real-life data in a compact state space, which is required for a successful verification. The resulting overflow probabilities of both the model and the empirical data are compared.