What are two possible ways to achieve to minimize the downtime of virtual machines (VMs) during maintenance periods?