An administrator needs to build a VM for a legacy, single-threaded application. According to best practice, how many virtual CPUs should the administrator use?

A single-threaded application is architecturally limited to one execution thread, so assigning exactly one vCPU matches the workload's actual parallelism. Adding extra vCPUs forces the hypervisor scheduler to find multiple physical cores simultaneously, increasing contention and scheduling latency without delivering any performance benefit to the application.

Two vCPUs waste scheduler resources because a single-threaded app cannot execute on both cores simultaneously and the hypervisor must still schedule both vCPUs on physical cores.

Four vCPUs amplifies scheduler overhead and resource contention on the host for a workload that is fundamentally incapable of parallel execution.

Eight vCPUs is severely oversized and would impose significant hypervisor co-scheduling complexity for an application limited to a single thread of execution.