What is a concern when implementing virtual switching in a hypervisor?

Virtual switches operate within the hypervisor, meaning all network traffic processing for virtual machines consumes host CPU cycles and memory bandwidth. High network I/O or a large number of VMs can lead to contention and limitations on the virtual CPU data bus, impacting network performance within the virtualized environment.

Virtual interfaces within a hypervisor do not have configurable duplex, speed, or negotiation settings in the same way physical network interfaces do; these are handled abstractly by the hypervisor's networking stack.

Virtual switches inherently support broadcast domain isolation through features like VLANs, and while misconfiguration can be an issue, the capability itself is a strength, not a fundamental implementation concern.

Packet forwarding across multiple virtual collision domains (which are standard for switched virtual networks) is a function of the virtual switch, and while performance is a consideration, the concern is the underlying resource constraints (like CPU) affecting that forwarding, not the existence of multiple collision domains itself.