par Latouche, Guy ;Nguyen, Giang
Référence Queueing systems, 80, 1-2, page (105-125)
Publication Publié, 2015-06
Article révisé par les pairs
Résumé : We extend to Markov-modulated Brownian motion (MMBM) the renewal approach which has been successfully applied to the analysis of Markov-modulated fluid models. It has been shown recently that MMBM may be expressed as the limit of a parameterized family of Markov-modulated fluid models. We prove that the weak convergence also holds for systems with two reflecting boundaries, one at zero and one at $$b >0$$b>0, and that the stationary distributions of the approximating fluid models converge to the stationary distribution of the two-sided reflected MMBM. In so doing, we obtain a new representation for the stationary distribution. It is factorised into a vector determined by the phase behaviour when the fluid is either at the level 0 or the level $$b$$b, and a matrix expression characteristic of the process when the fluid is in the open interval $$(0,b)$$(0,b).