LPAR

LPAR stands for logical partition. It is a form of hardware-level virtualization that divides a single physical server into multiple independent logical servers, called partitions. Each partition runs its own operating system instance and has access to a specified portion of the system’s CPUs, memory, and I/O resources. The partitions are isolated from one another, providing fault containment and separate administration.

In IBM systems, LPARs are created and managed by a hypervisor. On IBM Power Systems, the Power Hypervisor (PHYP) implements LPARs and works with PowerVM to allocate CPU, memory, and I/O to each partition. On IBM mainframes, the Processor Resource/System Manager (PR/SM) provides LPAR functionality. While the underlying technology differs, the core concept remains: multiple partitions share the same physical hardware while operating independently.

Resource management for LPARs includes assigning dedicated or shared CPU capacity, memory, and I/O adapters to each partition. Many implementations support dynamic or online reallocation of resources, enabling resize and workload rebalancing without powering down the system. I/O may be virtualized to allow partitions to access network, storage, and peripheral resources efficiently while maintaining isolation.

Operating system support typical for LPARs includes various UNIX-like systems such as AIX and Linux, as well as IBM’s own IBM i. LPARs are commonly used to consolidate workloads, enable testing and development environments, provide workload isolation, and support disaster recovery strategies, all while maintaining independent administration and security boundaries.

Overall, LPARs enable efficient utilization of powerful single servers by partitioning hardware resources into multiple, isolated operating environments.