Inactive (spare) PUs can be added concurrently, dynamically providing nondisruptive upgrade of
processing capability. They can be characterized, in any combination, as CPs, ICFs, IFLs, or
zIIPs.
For shared CP PR/SM configurations,
added CPs are brought into the pool of shared CPs, effectively increasing
the number of physical cores to be shared among partitions. To increase
the number of logical cores online to an LP, simply define the LP
with both initial and reserved logical cores. Then all you need to
do is configure on the extra logical core(s) after the concurrent
upgrade.
All models can define logical partitions with as many reserved
CPs as necessary. With thoughtful planning, there is never a need
for a disruptive increase in the number of logical cores.
The maximum initially online logical core width that logical partitions
can be defined and activated with at any point in time is as follows
(the maximum defined logical core width is as great as the total number
of CPs achievable with concurrent CPU upgrade):
Maximum initial logical cores defined for a dedicated partition equals physical
cores online for the current model minus physical cores
currently dedicated and online to other dedicated partitions minus the
maximum online number of shared CPs among all the activated logical
partitions that are using shared CPs.
Maximum initially online logical cores defined for a shared partition equals physical
cores online for the current model minus physical cores
currently dedicated and online to other partitions using dedicated
CPs.
Maximum total logical cores (including initial and
reserved) for any partition equals the number of physical
cores achievable through concurrent CP upgrade.
When a logical partition is defined to use ICFs, IFLs, or zIIPs these rules are applied against
the installed processors of that type. The total number of all logical processors defined to the
logical partition cannot exceed the maximum supported for a logical partition by the CPC,
independent of processor type.