2. PERFORMANCE REPORTING ANALYSIS › 2.6 PR/SM LPAR Performance Analysis › 2.6.2 PR/SM LPAR Concepts › 2.6.2.1 Overview of PR/SM LPAR Operation

2.6.2.1 Overview of PR/SM LPAR Operation

When an LPAR is defined, the amount of central and expanded
storage allocated to it is specified in one-megabyte
increments.  This storage is dedicated (not shared with nor
accessible in any way by other LPARs).

Individual central processors (CPs), on the other hand, can
either be dedicated to or shared among LPARs.  Dedicated CPs
are used exclusively by the LPAR and are not controlled in
any way by the PR/SM feature.  An LPAR can have multiple
logical CPs defined, but the number cannot exceed the number
of CPs under PR/SM control.  In other words, the total number
of CPs in the complex minus the number of dedicated
processors, if any, is the number of multiple logical
processors that can be defined to an LPAR.

PR/SM contains facilities to prioritize LPAR access to CPs
during periods of contention.  Within a single complex it is
possible to mix LPARs, including dedicated processors with
LPARs sharing processors, assuming sufficient CPs exist.  For
any LPAR with more than one defined logical processor,
however, all of them must be dedicated or all must be shared.

Vector processing elements are associated with the CPs to
which they are attached within the CEC, and can logically be
thought of as being a part of the CP itself for purposes of
this discussion.

The PR/SM feature allocates processor resources by assigning
CPs to logical processors as needed by the active LPARs.  If
the total demand of active LPARs is less than the total
resources of the complex, there is no limit placed on the
availability of CPs to any LPAR other than the limit implied
by the number of logical processors defined for the LPAR:
there can never be more CPs allocated than the number of
defined logical processors.  When demand exceeds supply,
PR/SM prioritizes CP allocations based on values specified
when the LPAR was defined (this weighting can be modified
dynamically by an authorized operator).  Depending on your
processor type, you may be able to cap LPARs to limit the
percentage of the total processor it uses.

CPs are allocated to an LPAR by means of time slicing.  The
duration of the time slice can be determined dynamically by
the PR/SM feature, and is based on the total number of
non-dedicated CPs and the number of active logical
processors.  The duration of a time slice can range from a
minimum of 12.5 milliseconds to a theoretical maximum of 300
milliseconds.  A CP will remain allocated to an LPAR until
either the time slice expires or the SCP running in the LPAR
attempts to place the logical processor in the wait state; at
this time, the CP will be made available for work elsewhere.