Abstract: As the rapid development of cloud computing techniques, a large number of data centers have been deployed recently. The power costs of data centers have become a practical issue and have attracted significant attention. Existing works on green data center have focused on computer servers and cooling systems. However, saving network energy also plays an important role on energy efficiency of data centers. In this study, we formally define the Network Power Saving VM Placement Problem (NPS-VMPP), analyze and highlight the resemblance between NPS-VMPP and a variant of Quadratic Assignment Problem (QAP) for Fat-Tree topology data centers. Inspired by this observation, we propose an extended robust tabu search based approach (eRTS) to optimize virtual machine placement and adopt a simple topology-aware heuristic to allocate traffic flow so as to turn off as many unneeded network devices as possible. Experiment results demonstrate the efficacy of this approach.