SUMMARYTraditional network dimensioning formulations have applied the Erlang model where the connections reserve capacity in the network. Until recently, tractable stochastic network models where the connections share the capacity in the network did not exist. The latter are becoming increasingly important as they can be applied to characterise file transfers in current data networks (e.g. IP networks). In particular, they can be utilised for dimensioning of networks with respect to the file transfer performance. To this end, we consider a model where the traffic consists of elastic flows (i.e. file transfers). Flows arrive randomly and share the network resources resulting in stochastically varying transmission rates for flows. Our contribution is to develop efficient methods for capacity planning to meet the performance requirements expressed in terms of the average transmission rate of flows on a given route, i.e. the per-flow throughput. These methods are validated using ns2 simulations. We discuss also the effects of access rate limitations and how to combine the elastic traffic requirements with those of real-time traffic. Finally, we outline how the methods can be applied in wireless mesh networks. Our results enable a simple characterisation of the order-of-magnitude of the required capacities, which can be utilised as a first step in practical network planning and dimensioning.