Infrastructure-as-a-service (IaaS) is an important type of cloud service. It integrates conventional computing resources, network resources, storage resources and related software packages into an on-demand network service, which provide virtual resources and machines to the users instead. Therefore, it is important to use a good VM placement mechanism that achieves good performance in an IaaS platform. For example, while allocating resources to a set of virtual machines (VMs), load balance is an important technique to avoid resource usage surge and thus can provide good quality of services to users. In this paper, we present a new VM placement mechanism based on resource usage prediction. The mechanism is implemented on a cloud platform, namely SAMEVED-Stack. The SAMEVED-Stack is a non-commercial cloud platform that supports high-level operations for virtual cluster/datacenter management. It is derived from OpenStack, and designed for supporting on-line network and computer security experiments. Our preliminary results show that, the proposed VM placement strategy outperforms three popular VM placement strategies, the Best-Fit, the First-Fit, and the max-free-memory-fit.