This paper presents an optimization framework for downlink transmission parameters of mobile cellular systems. A typical network optimization approach is to divide the network into disjoint clusters of base-stations (BS). Optimization is then performed within each cluster for important parameters such as transmit power, precoder, etc. This approach is widely adopted in academic research and industrial standard bodies, e.g. 3GPP LTE-Advanced. While reducing the optimization complexity, this strategy suffers from a performance limit due to interference from the nodes outside cluster. We thus propose a framework to overcome this limit by allowing clusters to exchange their parameters and optimization information via low-rate and non-zero delay backhauls. System-level simulations for LTE downlink transmit power optimization show that the proposed optimization model, while having the low-complexity of cluster-based approach, could nearly achieve the performance of network-wise optimization. Therefore, this model is particularly suitable for optimization of 4G and beyond-4G cellular networks.