System information

Chapter 1. Introduction to LPAR on IBM System i5 25
If you could determine the processor requirement for your partition workload and predict the
workload growth in the future that demands more processing power, we recommend that you
create a logical partition with dedicated processors for the best logical partition performance.
If you have a limited number of physical processors and want to create several logical
partitions with flexibility in processor resources usage among all logical partitions, then a
shared processor could be your choice.
1.3.12 Considerations in configuring virtual processors
The virtual processor is used to define a whole number of processors owned by a shared
logical partition from fractional processors assigned to it. The logical partition operations will
be faster and produce better results if the number of virtual processors assigned is very close
to the number of processing units from fractional processors assigned to that logical partition.
The addition of some virtual processors may slightly improve the performance of the
operating system because the operating system assumes that it has more processors to do
more concurrent jobs. However, adding too many virtual processors to a shared processor
logical partition will degrade the performance of each virtual processor. If you add too many
virtual processors for a logical partition and the processing unit is not increased, the
processing units for each new virtual processor will be smaller than for the previous virtual
processor. The operating system cannot shift processing powers among processes very well.
An uncapped shared logical partition can use idle processing units in the shared processors
pool. If the virtual processors for an uncapped shared logical partition is too low, it cannot use
the maximum processing units in the shared processors pool. Setting too many virtual
processors may impact logical partition performance as explained in the previous paragraph.
Determine carefully the proper number of virtual processors that will be configured for the
shared logical partition for optimum logical partition operations.
1.4 Memory use in System i5 LPARs
Memory on System i5 based System i5 systems is used and managed very differently than in
the past. In the following sections we discuss the new concepts of and changes to the
manipulation and use of memory.
1.4.1 Memory granularity
In the past we have been able to assign or move memory from a partition at the single MB
level. Memory in the System i5 machines cannot be manipulated that finely. Instead it must
be assigned or moved in blocks.
These blocks are referred to as Logical Memory Blocks (LMBs) but are more technically
described as Memory Regions. This is similar to the method by which the current Model 690
pSeries servers handle memory.
The LMBs are sized in multiples of 16 MB, and at initial announce, this will be the only size
that can be used. Later code updates will allow for use of 16, 32, 64, 128, and 256 MB blocks.
The designation of the memory region size you intend to use is a system wide parameter, so
that everything on a given system will be using the same size blocks.
What this means is that whatever LMB size is specified for the system, that will be the
smallest amount of memory that can be manipulated. All memory allocations will have to be in
multiples of the LMB size, and memory can only be moved between partitions at the LMB size
level.