ABSTRACT
Since the seminal work of Kelly and Roberts & Massoulie, bandwidth sharing networks are one of the predominant stochastic network tools to analyse congestion in communication networks. Recently, these networks have also been applied to analyze ramp metering in road traffic networks. From an OR perspective, these models are interesting, as their analysis requires both techniques from stochastic processes and optimization.
We will develop several approximations for the invariant distribution of such networks using fluid and diffusion scaling techniques. These approximations have the appealing property of being both insightful and computationally tractable.