As you will see in our “CPU & MEMORY RESOURCES” section, most of the world’s VMs are provisioned with more resources than they need to service their workloads — sometimes extensively so.
While compute over-provisioning can certainly result in waste in the private cloud data center — waste that can be reclaimed through rightsizing — there is another, more insidious way in which this over-provisioning can result in bad decision-making. This happens when organizations evaluate public clouds for possible workload migration.
On-premises, an over-provisioned VM may be hidden. This is because many system administrators manage resources at the cluster level, rather than focusing intently on the configuration of each and every VM. If an administrator is managing physical resources at the cluster level, then that administrator will continue to load VMs into the cluster until the overall cluster resources begin to be exhausted. Since the focus is on resource utilization, rather than resource configuration, the fact that VMs are over-provisioned can go unnoticed while the workload is on-premises.