Specifications
uses for nPartitions
One of the primary advantages of nPartitions is their support for flexible, user-defined partitioning
of resources. They also isolate hardware and software errors. Here are some of the typical types of
nPartitions:
• OS partitions—A partition can run various operating environments such as HP-UX, Windows NT,
and Linux. (Only HP-UX is available at first release.)
• OS version and functional partitions—Partitions on a system can run different versions of the same
operating system, and each can have a specific function (for example, a database partition, an
application partition, a development partition, a production partition, and so on).
• high-availability partitions—Partitions can participate in an MC/Serviceguard cluster, with
applications failing over from one partition to another.
• application partitions—A partition hosting a specific application can provide dedicated
resources, such as memory space.
• floating partitions—Applications will be able to take advantage of policy-based workload
balancing across partitions (not supported at first release).
• virtual partitions—Partitions can support workload consolidation and software isolation with
better granularity (for example, to the individual CPU level).
more details about some
Partitioning resources provides extraordinary flexibility in configuration and operation. Partitions
key uses for nPartitions
mean you can allocate physical resources and tune the operating system running on each
partition depending on the needs of the application (or the most important application) you intend
to run on it. Alternatively, you can configure the rp8400 as a single partition, allowing all the
resources to be focused on a single set of tasks—for example, a large online transaction-
processing application.
This section gives more details about some of the most popular and effective uses for nPartitions.
resource partitioning
Today, businesses require that multiple applications running on the same server deliver full
performance and do not conflict or interfere with one another. For top performance, it is critical
that the actions of one application not adversely impact or result in the denial of service to other
applications. Furthermore, a failure in the operating system or hardware components on the server
supporting a particular set of applications must not result in the loss of multiple applications
running on that server.
In order to have effective consolidation of applications on a single server, an application must not
be allowed to use a disproportionate share of system resources needed by another application,
and the failure of a hardware component must not result in the loss of service for multiple
applications. While a solution to these problems could be accomplished by providing multiple
servers at a site, such a method is inflexible in the face of changing requirements, is more
expensive, and requires more space in the data center. Instead, the IT department needs a flexible
configuration that offers the capability to adjust compute resources in a dynamic fashion while still
preserving the symmetric multiprocessor (SMP) programming model.
That’s why nPartitions are so effective at resource partitioning. With rp8400 nPartitions, each
partition is isolated from the other with special, dynamically configurable hardware that provides
an effective firewall between applications. There is no possibility of interference between
nPartitions.
With nPartitions, the system manager has the control necessary to optimize the allocation of
resources to competing workloads. For example, as workloads change through the daily cycle,
the manager can deploy system resources to optimize interactive performance applications during
normal business hours. When peaks of interactive activity occur, additional resources can be
added to cover the load. As the interactive load diminishes, resources can be returned to favor
batch or other workloads. At any time, the administrator can select cells and move them from one
partition to another in order to support resource requirements in a specific partition. And the
movement of resources can even be accomplished automatically through the use of control scripts.
25










