System information

Chapter 1. Introduction to LPAR on IBM System i5 11
However, job B1 must stop after 2 milliseconds in the second processor cycle again because
the processing units assigned are very small. In the next processor cycle, job B1 is resumed
again, and has to stop after 2 milliseconds. In the next processor cycle, job A2 is submitted
and finishes after 6 milliseconds. Jobs B1 is resumed again, being processed for 2
milliseconds, and is complete.
This example shows two logical partitions sharing a single processing unit from a shared
processors pool. A logical partition with adequate processing units assigned to it will have its
jobs complete faster than if it had smaller processing units. It is very important to determine
the processing units required by a logical partition for better logical partition performance.
Simultaneous Multi Threading (SMT)
System i5 Processors implement Simultaneous Multi Threading (SMT) technology to process
the instructions given. The instructions from the operating system are loaded simultaneously
into the processor and executed. When the instructions are loaded simultaneously into the
processor, they have to wait their turn for registers that needed by each instruction at the
same time. This makes total time longer to execute each instruction, but total time to execute
all the instructions loaded simultaneously is shorter. In this case, the SMT technology
implemented in the System i5 processor results in greater system capacity because multiple
instructions can be executed at shorter time.
During periods of low user requests, instructions are infrequently executed and reach the
processor to be executed. In this period, there will be only one thread in the processor pipes
at any point in time. The absence of contention allows the processor to complete each
instruction faster. When demand for high processing power arises due to an increase in the
frequency of user requests, SMT technology allows greater parallelism in the processing of
instructions and reduces overall elapsed time for multiple requests.
The Simultaneous Multi Threading (SMT) is controlled by the QPRCMLTTSK system value of
the OS/400 operating system. Each operating system in the logical partitions can define its
own QPRCMLTTSK system value independently. The logical partition must be restarted in
order to apply changes in QPRCMLTTSK system value.
There are three available values to set the SMT in QPRCMLTTSK system value. Value ‘0’ will
turn off SMT; the result is that there will be only one task per physical processor. Value ‘1’ will
turn on SMT and make two tasks available per physical processor. Value ‘2’ will turn on SMT,
but the operating system might switch to single thread mode.
When SMT is turned on, up to two threads are dispatched to each physical processor, thus
resulting in an environment with two logical processors. If SMT is turned off, only one thread
will be dispatched and the system will only operate in an environment with one logical
processor. When the value of ‘2’ is given to QPRCMLTTSK, which means that the SMT
environment control is given to the system, the system will monitor the rate of instruction
execution requests to the processor and switch between the two environments to achieve the
best system performance.
1.3.2 Logical partition resources
Logical partitioning on the System i5 system allows us to create several independent logical
partitions with their own processors, memories, and I/O resources. The logical partition profile
defines these resource allocations for each logical partition. Each logical partition can run
independently within the System i5 system.
The Power Hypervisor allocates the Server processors and memory during startup and
during dynamic reallocation of processor and memory using Dynamic Logical Partitioning
(DLPAR). Each operating system (i5/OS) allocates the processors and the memory available