Partitioning refers to the division of one server into multiple processing units. This allows the OS and applications to run in separate independent areas. In this way, multiple systems can be built within a single server. Each individual area is called a "partition."

PRIMEPOWER supports highly reliable and flexible "hardware partitioning" and "extended partitioning." Each partition is configured in units of a system board that has CPU and memory mounted on it. Each system board is physically independent from other boards. Even if the load in one partition increases or problems arise, all other partitions are unaffected.

This feature is found in PRIMEPOWER 900 and above.

Hardware Partitioning (PPAR: Physical Partitioning)

Hardware partitioning divides hardware resources in systemboard units. The installed CPU and memory, allows the OS and applications to be run on each hardware sub-division. . Everything from hardware resources to OS and applications can be fully divided for each system.

Since this technology involves partitioning by system board units, fewer partitions can be created with this technology. But this technology offers the advantage that even if a failure occurs in hardware, the effects are confined to only one partition, and only the application system associated with the point of failure is affected. This securely isolates failures to only one partition. Since partitions are controlled by hardware, there are no adverse performance drawbacks attributable to such partitioning.

Extended Partitioning (XPAR: eXtended Partitioning)

In addition to systemboard level hardware partitioning, PRIMEPOWER offers Extended Partitioning (XPAR), which combines the highly reliable partitioning function with flexibility.

To maximize resource power while retaining the failure-isolation of hardware partitioning, PRIMEPOWER allows the system board to be divided into four (PRIMEPOWER 900 and 1500) or two (PRIMEPOWER 2500) CPU partitions. This "Extended Partitioning" enables a more elaborate system design close to logical partitioning, while maintaining the high reliability of hardware partitioning.

Each partition can run a different version of Solaris™.

Since the partitions are completely independent, a failure in one partition will never propagate to other partitions. Compared to software partitioning or logical partitioning, extended partitioning offers high failure isolation capability, with few locations where a failure at one point can bring down the entire system. This helps ensure uninterrupted business operations.

To help protect client IT investments, PRIMEPOWER allows CPUs operating at different clock speeds to coexist on a system board. What's more, clients can combine CPUs operating at different clock speeds within a single partition. This means that PRIMEPOWER can begin with a single system board.

When additional resources are needed, users can install additional system boards with faster CPUs.

Since the latest-spec CPUs can be installed when needed, the system can be expanded flexibly to suit the type and scale of a particular user's business operations or strategies.

Since a system can be built by dividing everything from the hardware resources to the OS and applications, it's possible to consolidate on a single server a system that would otherwise consist of multiple small rack-configured servers.

And while systems comprised of small servers cannot share hardware resources, those built with a single server based on partitioning permit the sharing of unused server resources between systems.

Dynamic Reconfiguration (DR)

Hardware resources such as CPU, memory, and I/O can be moved from one partition to another while the system remains active, even when operating in divided partitions. This feature is called Dynamic Reconfiguration (DR). Since hardware resources can be incorporated into or removed from partitions dynamically, the system can be reconfigured as necessary.

Dynamic Reconfiguration makes it possible to add or distribute resources to or between systems based on operational demands. Dynamic Reconfiguration allows flexible use of PRIMEPOWER capacity (expandable up to 128 CPUs), making it ideal for server consolidation and other such applications.

Once spare resources are prepared, it is possible to incorporate those spare resources into systems based on daily, monthly, or yearly changes in business operations load. In systems that must operate 24x7, dynamic reconfiguration helps implement flexible resource allocation based on changing data volumes or business operations load.

When partitions are created using hardware resources on multiple system boards, if a certain CPU in one of those partitions fails, causing the system to degenerate, the defective CPU can be separated dynamically and exchanged with another CPU without halting the entire system. The new CPU can then be incorporated dynamically into the partition in which the original was located.