118
IBM Power 595 Technical Overview and Introduction
that these maximums are supported by the hardware, however the practical limits depend on
the application workload demands.
The processing resources allocated to an SPLAR are called processing units (PU). The PU
setting represents a guaranteed amount of processing power that is available to a given
SPLPAR no matter how busy the shared processor pool is.
If the SPLAR is configured as
uncapped
, it will have the ability to access excess processing
units above is specified PU setting. If needed, the PU setting can be changed using a
dynamic LPAR operation. The shared processor pool, SPLARs, and PUs are shown in
Figure 3-1 on page 117.
Shared processor pool resources are abstracted to the SPLPARs through virtual processors.
Virtual processors (VPs) map the processing power (PUs) assigned to the operating system
running on the SPLPAR. VPs are specified in whole numbers.
The number of VPs supported on SPLPAR depends on the number of PUs assigned to that
SPLPAR. Each VP can represent from 0.1 to 1 processing units. The minimum number of
VPs allowed is equal to the PUs assigned rounded up to the nearest whole number. The
maximum number of VPs allowed is equal to the assigned PUs multiplied by ten. The
assigned number of VPs sets an upper limit for an SPLPAR ability to consume excess
processor units. For example, a SPLPAR with three VPs can access up to three processing
units. The number of VPs can be changed dynamically through a dynamic LPAR operation.
Every ten milliseconds, the POWER Hypervisor recalculates each SPLAR’s processing
needs. If a SPLAR is not busy, it receives an allocation that is smaller than the assigned
processing units, making this SPLAR a
donor
of processor cycles to the shared processor
pool. If the SPLPAR is busy (and
uncapped
) the Hypervisor can assign excess processor
cycles to the SPLPAR up to the number of VP assigned.
3.5.3 Shared dedicated capacity
On POWER6 servers, you can now configure dedicated partitions to donate excess
processor cycles to the shared processor pool when their assigned processors are idle. At all
times, the dedicated partition receives priority access to its assigned processors. Enabling
the shared dedicated capacity feature can help to increase system utilization, without
compromising the computing power for critical workloads running in partition with dedicated
processors.
3.5.4 Multiple shared processor pools
The original shared processor pool implementation provides support for one shared
processor pool. Starting with POWER6, multiple shared processor pools can be defined.
SPLARs can now be assigned to run in a specific pool. Each pool has a maximum processing
unit (MaxPU) setting that limits the total amount of processing units available to the collection
Note:
The sum total of the PU settings for all running SPLPARs within a 595 server must
be less than or equal to the size of the shared processor pool. This ensures that all
SPLARs can simultaneously access their guaranteed amount of processing power.
Note:
Your 595 server is positioned to begin sharing physical processors when the total
number of running VPs exceeds the size of the shared processor pool.
Summary of Contents for Power 595
Page 2: ......
Page 120: ...108 IBM Power 595 Technical Overview and Introduction...
Page 182: ...170 IBM Power 595 Technical Overview and Introduction...
Page 186: ...174 IBM Power 595 Technical Overview and Introduction...
Page 187: ......
