Optimizing energy consumption is paramount to sustain the growth of cellular networks. One of the approaches to reduce energy consumption is traffic dependent operation of networks. The traffic demand experienced by the network fluctuates over the duration of a day. Therefore, during the periods of low traffic, we may re-configure the network to trade the excess capacity for energy reduction. For example, we may modulate the BS transmit power (Cell Zooming) so as to maintain the desired QoS. However, determining an optimal network configuration is known to be computationally hard. Along with Cell Zooming, it is essential to also consider channel re-assignment, for, when the transmit powers of BSs are changed, the channels allocated to BSs must also be suitably changed. Considering therefore channel allocations also as state variables, the search space over which the optimization should be performed blows up, further complicating the problem. In this paper, we propose a framework to address this problem. The proposed algorithm is suitable for such large search spaces, while the framework is general enough to admit several QoS requirements and sophisticated power consumption models. The underlying mathematical formulation is also applicable in other contexts, and is of independent interest as well.