The robustness and integrity of IP networks require efficient tools for traffic monitoring and analysis, which scale well with traffic volume and network size. We address the problem of optimal large-scale flow monitoring of computer networks under resource constraints. We propose a stochastic optimization framework where traffic measurements are done by exploiting the spatial (across network links) and temporal relationship of traffic flows. Specifically, given the network topology, the state-space characterization of network flows and sampling constraints at each monitoring station, we seek an optimal packet sampling strategy that yields the best traffic volume estimation for all flows of the network. The optimal sampling design is the result of a concave minimization problem; then, Kalman filtering is employed to yield a sequence of traffic estimates for each network flow. We evaluate our algorithm using real-world Internet2 data.