DQDB MODELING - PROBLEM COMPLEXITY REDUCTION AND SOLUTION VIA MARKOV-CHAINS

This paper presents a novel approach to DQDB modeling; the innovative aspect lies in the attempt to provide an accurate representation of the process of the busy slots travelling on the forward bus. We show in the paper that for aggregate loads exceeding 80% of the medium capacity, the Bernoulli hypothesis generally used for modeling the length of busy trains diverges greatly from the actual behavior. In this paper, by assuming that consecutive busy trains form a sequence of i.i.d. random variables, we compute the busy train statistics via a Markov chain approach. The results obtained show a marked improvement, compared to the Bemoulli hypothesis. Using our busy train statistics we have obtained an approximation of the bus access delay for each node of a DQDB network1.