52 IBM System p5 520 and 520Q Technical Overview and Introduction
A partition can be defined with a processor capacity as small as 0.10 processing units. This
represents one-tenth of a physical processor. Each physical processor can be shared by up
to 10 shared processor partitions, and a partition’s entitlement can be incremented
fractionally by as little as one-hundredth of the processor. The shared processor partitions are
dispatched and time-sliced on the physical processors under control of the POWER
Hypervisor. The shared processor partitions are created and managed by the HMC or
Integrated Virtualization Management (included with Virtual I/O Server software version 1.2 or
later). There is only one pool of shared processors at the time of writing this publication and
all shared partitions are dispatched by Hypervisor within this pool. Dedicated partitions and
micro-partitions can coexist on the same POWER5+ processor-based server as long as
enough processors are available.
The systems support up to a 4-core processor configuration, therefore, up to four dedicated
partitions, or up to 40 micro-partitions can be created. It is impor tant to point out that the
maximums stated are supported by the hardware, but the practical limits depend on the
application workload demands.
2.12.2 Logical, virtual, and physical processor mapping
The meaning of the term physical processor in this section is a processor core. For example,
in a 2-core server with a DCM (dual-core module), there are two physical processors, and in a
4-core configuration with a QCM (quad-core module), there are four physical processors.
In dedicated mode, physical processors are assigned as a whole to partitions. The
simultaneous multithreading feature in the POWER5+ processor core allows the core to
execute instructions from two independent software threads simultaneously. To support this
feature, the concept of logical processors was introduced. The operating system (AIX 5L or
Linux) sees one physical processor as two logical processors if the simultaneous
multithreading feature is on. It can be turned off while operating system is executing (for AIX
5L, use the smtctl command). If simultaneous multithreading is off, then each physical
processor is presented as one logical processor, and, thus, only one thread is executed on
the physical processor at the time.
In a micro-partitioned environment with shared mode partitions, an additional concept of
virtual processors was introduced. Shared partitions can define any number of virtual
processors (maximum number is 10 times the number of processing units assigned to the
partition). From the POWER Hypervisor point of view, the virtual processors represent
dispatching objects (for example, the POWER Hypervisor dispatches virtual processors to
physical processors according to partition’s processing unit entitlement). At the end of the
POWER Hypervisor’s dispatch cycle, all partitions should receive total CPU time equal to
their processing unit entitlement. Virtual processors are either running (dispatched) on a
physical processor or standby (waiting). The operating system is able to dispatch its software
threads to these virtual processors and is completely screened from the actual number of
physical processors. The logical processors are defined on top of virtual processors in the
same way that physical processors are defined. So, even with a virtual processor, the
concept of logical processor exists and the number of logical processors depends on whether
the simultaneous multithreading is turned on or off.
Some additional information related to the virtual processors:
There is a one-to-one mapping of running virtual processors to physical processors at any
given time. No more virtual processors can be active at any given time than the total
number of physical processors in the shared processor pool.
A virtual processor can be either running (dispatched) on a physical processor or standby
waiting for a physical processor to become available.