A lot of attention has been devoted to flow routing in networks by the past. Usually, each commodity to be routed is characterized by a single deterministic traffic value. However, telecommunication traffic is well known to be highly variable in time. The current paper revisits usual network routing models, in considering traffic stochasticity, with the aim of better modeling traffic multiplexing. Appropriate stochastic optimization models for performance-oriented routing are proposed. Numerical tests illustrate the gain of routing according to our stochastic model, when compared to other more usual routing approaches.