Abstract-We study a multi-node network, where a multi-antenna transmitter Tx communicates with its desired receiver Rx, while a cluster P {Px,n, n = 1, . . . , N } of unintended nodes is disturbed by the Tx-Rx (TR) communication. To prevent severe performance degradation, we impose a constraint on the total interference that is inflicted at the nodes of P. The TR link contains a line-of-sight component whereas the propagation environment for each Tx-Px,n link is shadowed. The Tx node is preprocessing the information sequence by means of a precoding matrix that is optimized to achieve the ergodic capacity under a constraint on the maximum admissible ergodic interference power, arriving on P. In this paper, we show that the optimum precoding strategy involves transmission of a single stream over the precoding direction, i.e., the eigenvector of the precoding matrix, which corresponds to beamforming along the instantaneous direction of the TR link channel. Solution of the remaining power allocation problem yields the optimal precoding matrix. For this setup, we provide an efficient stochastic characterization of the network, that allows us to obtain an analytical expression for the TR link ergodic-capacity; this problem has been previously open even for the case of a single-antenna node Tx and a single-element set P. We complement the analysis by deriving the TR link signalto-noise ratio and the average bit-error-rate, which are associated with our transmission scheme. Numerical results corroborate the theoretical analysis and reveal an interplay between the network parameters and their impact on the TR link performance.