Performance on a new private cloud deployment is not meeting expectations. An engineer has determined the virtualization host is experiencing high ready/wait times. VMs with the following…

Decreasing the vCPU count for VM1 reduces that VM's scheduling footprint but does not address the memory pressure on VM2 that is generating the CPU overhead contributing to host-wide ready/wait times.

When a VM is starved of memory, the hypervisor invokes memory reclamation techniques such as balloon drivers or host-level swap, both of which generate significant CPU overhead and cause vCPUs to stall waiting for memory operations to complete. The metrics for VM2 indicate it is memory-constrained, so increasing its memory allocation eliminates the pressure driving these reclamation cycles. This removes the root cause of the elevated ready/wait baseline rather than simply redistributing the contention.

Decreasing memory for VM3 would intensify memory pressure on that VM, increasing balloon and swap activity and worsening CPU overhead across the host rather than improving it.

Increasing the vCPU count for VM4 adds more virtual CPUs competing for the same physical cores, which raises CPU scheduling contention and directly increases ready time rather than reducing it.