Stationary distributions and convergence for \(M/M/1\) queues in interactive random environment (Q2306745): Difference between revisions

For a classical \(M/M/1/\infty\) queueing system let the arrival rates \(\lambda(z)\) and service rates \(\mu(z)\) (which are queue length independent) depend on an external parameter \(z\in D\), where \(D\) is called the environment of the queueing system. The main results: Stationary distribution (if it exists) of the coupled ``interactive'' system queue-environment and speed of convergence to equilibrium. The term ``interactive'' emphasizes the property that not only the transition rates of the queue depend on the status of the environment: The dynamics of the environment depends on the queue length as well. So, the environment process is not Markov. The environment's state space is either a countable set \(D\) (as it can be found in the literature as standard up to now), i.e., the dynamics of the environment constitute a pure jump process, or \(D\) is a subset of \(\mathbb{R}^n\), and the environment is a reflected jump diffusion. In both cases the transition mechanism of the environment is constructed in a special way such that the stationary distribution is found to be of simple ``separable'' form: The two-dimensional stationary distribution at any fixed time instant is the product of its marginals (in space), i.e., product form equilibrium.