A benchmark model for parallel ns3

ns3 is a simulation framework for computer networks, derived from a long line of serial simulators. Recently, ns3 incorporated a parallel, distributed scheduler, which enables distributed ns3 simulation for the first time. In this paper we discuss the current implementation and some of its limitations, with an eye to exploring potential improvements. In order to gauge progress, it is essential to have a meaningful performance metric and a suitable benchmark problem. Therefore we outline how to measure the simulation critical path and use that to construct a parallel performance metric. Second, we propose a scalable benchmark model, inspired by the global Internet.