We consider the problem of designing transmit and receive precoders to minimize the number of backlogged packets at the base stations (BSs) in a multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) scenario. Due to the nonconvex nature of the problem, we propose an iterative method for designing transmit precoders and receive beamformers. Since the overhead involved in exchanging channel state information (CSI) to a centralized controller is large, we propose a distributed precoder design using Karush-Kuhn-Tucker (KKT) expressions. In order to efficiently implement precoders and to reduce the signaling overhead, we use bi-directional training (BiT) design to feedback the involved variables in an analog format. By doing so, we penalize the achievable throughput by the number of resources consumed in the BiT procedure. A realistic dynamic traffic model is considered, where the transmission rates of all users depend on the current arrivals and the number of backlogged packets present in the respective BS. To improve the performance with limited BiT steps, we perform BS specific updates for the corresponding transmit and receive beamformers between each BiT exchange. Numerical simulations are provided to highlight the benefits of proposed schemes by considering multiple transmission slots under time-correlated fading model.