According to a computer-implemented method, a first virtual machine (VM) is deployed on a first hypervisor from a non-clustered server pool to run a workload of one or more applications. A dummy VM is configured on a second hypervisor from the non-clustered server pool to reserve same resources as the first VM without powering the dummy VM. The first VM is powered with a cold start on the second hypervisor using the resources on the dummy VM. Also, the first VM is provided with a same VM configuration on the second hypervisor that was on the first hypervisor.