QRP03-4: Efficient Simulation of Bandwidth Allocation Dynamics in P2P Networks

Fluid-flow simulation models of peer-to-peer (P2P) networks are an effective way to properly account for low-level network bottleneck details, meanwhile circumventing the scalability limits imposed by packet-based simulators. A fundamental component in a fluid-flow simulator is the algorithm for the computation of the achievable per-flow rate for several competing flows offered to a given network topology. This paper presents a novel efficient approach to compute, in an exact manner (i.e. with no approximations), the max-min fair rate allocation. Numerical results demonstrate that our proposed algorithm outperforms traditional max-min computation approaches by as much as a factor 100 for a million nodes network.