PRM Product Overview

10
groups. You can be assured that critical applications will get the real memory they need to meet their
service level objectives.
The PRM memory manager enables you to assign shares (a guaranteed minimum allocation) of
private real memory. PRM uses a kernel-space method for enforcing minimum memory allocations.
PRM pages out the memory of applications in PRM groups that are consuming more than their
configured memory allocations. When excess memory is available, sharing of the excess is the default
behavior. However, you can restrict which groups can borrow or lend their memory resources,
potentially isolating memory for your groups. This isolation can be particularly beneficial when
paging out memory is time-intensive. Memory capping is also available. When a PRM group’s
memory is being isolated or capped, if the group attempts to exceed its minimum memory allocation,
PRM frees more memory by paging out the memory used by that PRM group.
Memory shares can be changed dynamically while the system is in operation to reflect changes in
application priorities.
Combining memory management with pSet PRM groups, you can completely isolate a group’s CPU
and memory resources, which prevents other PRM groups from accessing that group’s CPU or
memory resources.
Managing shared real memory allocation
By default, all shared memory is allocated in the PRM_SYS group.
Starting with HP-UX 11i v2 Update 2 and PRM C.03.01, PRM can control shared memory allocations
on a PRM group basis. You only control shared memory for the groups that need it—you can omit
control for groups where shared memory control would not be helpful.
You set a minimum size in megabytes for a group’s shared memory allocation. (This size is usually
available from the configuration settings for the consuming application, as is the case with the Oracle
SGA size.)
Capping is not available for shared memory.
Managing disk I/O bandwidth allocation
Applications can underperform if they cannot complete their disk reads and writes in a timely manner.
Hence, it is important to treat disk I/O bandwidth as a critical resource and manage it to ensure
critical applications do not underperform because of contention for disk bandwidth.
PRM provides a disk I/O bandwidth control feature that enables you to define the priorities of disk
read and write operations for different PRM groups. The PRM disk I/O control works with the HP
Logical Volume Manager (LVM) and the VERITAS Volume Manager (VxVM). The I/O control operates
at the volume group layer (with LVM) or disk group layer (with VxVM) in the operating system kernel.
PRM controls disk bandwidth usage by reordering the I/O requests of a volume group (disk group).
When the volume group (disk group) is busy, requests from low-priority processes are delayed and
requests from high-priority processes are accelerated. Operating at this layer provides the capability
to control the widest array of disk technologies in a single, easy-to-configure feature.
The disk I/O bandwidth shares can be modified while the volume group (disk group) is in use.
Unused disk I/O bandwidth is shared among active PRM groups in proportion to their PRM disk I/O
shares.
PRM cannot manage disk I/O bandwidth for pSet PRM groups.