Sojourn times in fluid queues with independent and dependent input and output processes

Abstract

Markov Fluid Queues (MFQs) are the continuous counterparts of quasi birth–death processes, where infinitesimally small jobs (fluid drops) are arriving and are being served according to rates modulated by a continuous time Markov chain. The fluid drops are served according to the First-Come–First-Served (FCFS) discipline. The queue length process of MFQs can be analyzed by efficient numerical methods developed for Markovian fluid models. In this paper, however, we are focusing on the sojourn time distribution of the fluid drops.In the first part of the paper we derive the phase-type representation of the sojourn time when the input and output processes of the queue are dependent. In the second part we investigate the case when the input and output processes are independent. Based on the age process analysis of the fluid drops, we provide smaller phase-type representations for the sojourn time than the one for dependent input and output processes.