We consider the problem of minimizing the routing power consumption in software-defined networks. The network is composed of software-defined networking (SDN) nodes and a central controller where routing decisions are centralized. More specifically, the central controller minimizes the routing power consumption by routing flows on the minimum number of active links with the lowest discrete link rates. Thus, it maximizes the number of inactive links and the level of link rates, which reflects significant saving in power consumption. This problem is a mixed-integer programming problem and known to be NP- hard. Therefore, we propose a low-complexity greedy heuristic to minimize the number of active links and link rates by rerouting flows and aggregating them on common links. Numerical results show that the proposed algorithm achieves 17.18% to 32.97% power saving in real network topologies relative to a base shortest path algorithm. The savings are achieved with minimal increase in average path length that is less than 0.2 hops.