The popularity of cloud computing has led cloud provider to increase the number of data center to handle the growing clients demand. This trend has inevitably lead to huge electricity consumption. However due to the nature of cloud computing, resources are controlled by the cloud provider and thus can be managed in a more efficient manner than physical resources. In Infrastructure-as-a-Service (IaaS) clouds, clients execute virtual machines (VMs) which are distributed across many nodes. One way to reduce the electric consumption is to migrate VMs on as fewer nodes as possible and then, put idle nodes in a lower power state. This problem is an instance of the bin-packing problem (BPP) which is known to be NP-hard, therefore recent algorithms focus on heuristic methods for finding near optimal solutions in a feasible amount of time. Current existing solutions are either centralized which limits scalability either hierarchical which lead to local optimizations or either use ring strucured P2P solution which is hard to maintain and need in worst cases full traversal of the ring to find nodes. We propose a scalable fully distributed energy aware epidemic protocol to optimize the placement of VMs while taking into account the number of migrations.