Para-partitioning is an Extensible Firmware Interface-based (EFI) mechanism to create multiple logical systems on a single physical system. While it is comparable to virtualization in functionality, para-partitioning does not virtualize memory, devices or any other resources. All resources are partitioned amongst logical instances at boot time, and an OS in each logical system receives resources allocated to it. These schemes are not mutually exclusive, and can coexist.

Para-partitioning and virtualization have different tradeoffs. In para-partitioning, there is no hypervisor, no abstraction or emulation layer slowing down access to physical devices. However, resource mapping is strictly 1:1, including processor cores, and the OSes must be well-behaved to ensure that they only access their own devices. Virtualization's hardware abstraction allows hypervisor-protected sharing of physical devices and can support more guest OSes than physical processor cores.

With virtualization, guest OSes see virtual processors; each guest OS may simply run as a single thread on the host, limiting the overall performance of each guest regardless of the number of processing cores available. With para-partitioning, however, specific cores can be dedicated to specific guests. Each guest's user processes can be scheduled on multiple cores instead of sharing the guest's virtual execution thread.

The primary application areas targeted with para-partitioning are network/telecom applications, consolidation of RTOS-based or DSP applications, and to facilitate scaling of SMP-limited applications.