Partition Load Manager Standards
The Partition Load Manager (PLM) provides CPU and memory
resource management and monitoring across logical partitions (LPARs).
Partition Load Manager allows you to effectively use CPU and Memory
resources by allowing you to set thresholds for designated resources.
When a threshold is exceeded, Partition Load Manager can try to assign
CPU and/or Memory resources to that LPAR by using resources assigned to
other LPARs that are not being used.
Determining which node is more or less deserving of resources is
primarily done by taking into account certain values defined in what is
known as a policy file. This policy file details partitions, their
entitlements, their thresholds, and organizes the partitions into
groups. Every node, but not every LPAR, managed by Partition Load
Manager must be defined in the policy file along with several associated
attribute values. Some of the attributes that are associated with the
node are the maximum, minimum, and guaranteed resource values, variable
share values, and so on. These are the attributes taken into
consideration by Partition Load Manager when a decision is made as to
whether a resource is reallocated from one LPAR to another.
PLM is an automated mechanism for utilizing the Dynamic LPAR (DLPAR)
capabilities of the HMC and requires communication with the HMC. This
means that before PLM will function, DLPAR must be functional on the
HMC. DLPAR requires communication with each LPAR via the Resource
Monitoring and Control (RMC) subsystem.
NOTE: The RMC subsystem is not installed when the
AIX operating system is installed from the NIM server as an "rte"
The following fileset must be installed on every PLM client LPAR to
enable RMC communications with the HMC and PLM:
The PLM communications are also dependent upon SSH and SSL and must
be installed on every PLM client LPAR.
Refer to the PLM configuration
procedures for more information
A single PLM server can manage multiple frames across multiple HMC's.
In the Mt Xia environment there is a single primary PLM in each data
center. Within a frame there are two classifications of CPU's,
dedicated and shared. Policy files are used by the PLM to control each
frame and a single policy file will exist for each frame. The policy
file is named for the serial number of each frame. When new frames are
added to Mt Xia's environment, a policy file will be created on the PLM and
the name of the policy file will be the serial number of the frame.
Policy files currently exist with names such as:
- 0066ACAH - p550 - Mt Xia
- 01234567 - p570 - Mt Xia
Every LPAR created in the Mt Xia environment will be managed by a PLM
and will be initially assigned a minimum amount of CPU and memory
resources. This means there will be a policy file on the PLM for every
frame in the data center where the PLM exists.
Within a PLM policy there are two groups to represent the two CPU
classifications, dedicated and shared. Each LPAR will be assigned to one
of these two groups, depending upon what type of CPU's are assigned to
As a configuration standard, every policy will be configured to
immediately release free CPU and memory resources. Most other
configuration parameters within the PLM will depend upon the LPAR and