Micro-Partition Overview

© Copyright IBM Corporation 2006 Course materials may not be reproduced in whole or in part without written permission of the AIX Collaboration Center
3.2
Shared Processor LPARs – Micro-partitions
IBM System p
„ LPARs are defined to be dedicated or shared
„ Dedicated partitions use whole number of CPUs

„ Shared partitions use whole or fractions of CPUs from Shared Pool

„ Shared processor pool - subset of physical CPUs in a system

„ all CPUs that are not in dedicated LPARs

„ Processing capacity for a micro-partition specified in terms of processing units.

„ The smallest capacity that can be specified for a partition is 0.1 Processing units,

equivalent to 1/10 of a processor

„ Additional processing capacity can be configured in fractions of 0.01 processing units.

„ Configure the following options for processing capacity:

„ Desired: Size of partition at activation, between minimum and desired
„ Minimum: Partition won’t start if Minimum capacity not available
„ Maximum: CPU that can’t be exceeded in DLPAR operation

„ Capped vs uncapped
„ Capped: CPU Capacity limited to ‘desired’ entitlement
„ Uncapped: CPU Capacity limited by unused capacity in ‘pool’

„ LPAR weighting to determine preference when pool cycles are constrained

„ Some LPARs more favored (up to weight 255), some less favored (down to weight 1)

© Copyright IBM Corporation – AIX Collaboration Center 2006

Processor terminology
IBM System p

Shared processor Shared processor Dedicated

partition partition processor partition Logical (SMT)
SMT Off SMT On SMT Off
Virtual

Shared

Dedicated

Inactive (CUoD)
Entitled capacity

Deconfigured

Installed physical
processors

Shared processor pool

© Copyright IBM Corporation – AIX Collaboration Center 2006

Understanding min/max/desired resource values
IBM System p

• The desired value for a resource is given to a partition

if enough resource is available.

• If there is not enough resource to meet the desired

value, then a lower amount is allocated.

• If there is not enough resource to meet the min value,

the partition will not start.

• The maximum value is only used as an upper limit for

dynamic partitioning operations.

© Copyright IBM Corporation – AIX Collaboration Center 2006

Partition capacity entitlement example
IBM System p
• Shared pool has 2.0 processing units available
• LPARs activated in sequence
• Partition 1 activated
- Min = 1.0, max = 2.0, desired = 1.5
- Starts with 1.5 allocated processing units
• Partition 2 activated
- Min = 1.0, max = 2.0, desired = 1.0
- Does not start
• Partition 3 activated
- Min = 0.1, max = 1.0, desired = 0.8
- Starts with 0.5 allocated processing units

© Copyright IBM Corporation – AIX Collaboration Center 2006

Capped and uncapped partitions
IBM System p

• Capped partition
- Not allowed to exceed its entitlement
• Uncapped partition
- Is allowed to exceed its entitlement
• Capacity weight
- Used for prioritizing uncapped partitions
- Value 0-255
- Value of 0 referred to as a “soft cap”

© Copyright IBM Corporation – AIX Collaboration Center 2006

 IBM System p

Capped Shared Processor LPAR

Pool Idle Capacity Available

Maximum Processor Capacity

Processor Entitled Processor Capacity

Capacity
ceded
LPAR Capacity capacity
Utilization
Utilization
minimum processor capacity

utilized capacity

Time

© Copyright IBM Corporation – AIX Collaboration Center 2006

 IBM System p

Uncapped Shared Processor LPAR

Pool Idle Capacity Available

Maximum Processor Capacity

Processor
Capacity
Entitled Processor Capacity
Utilization
ceded capacity
minimumprocessor capacity

Utilized Capacity

Time
© Copyright IBM Corporation – AIX Collaboration Center 2006
Virtual Processors
IBM System p

virtual CPU
zShared processor concepts splpar 2
¾partitions run on virtual virtual timebase
processors
¾Partition entitled capacity are
divided among the virtual Dispatch
virtual CPU
virtual CPU virtual CPU
processors in the LPAR splpar11
splpar splpar 3
¾Virtual processors are virtualtimebase
timebase
Wheel (10ms) virtual timebase

virtual virtual timebase

dispatched on physical
processors
virtual CPU
splpar 4
virtual timebase
virtual timebase
Dispatch Window
0 1 2 3 4 5 6 7 8 9 10
Dispatched
Physical
processor
physical CPU
1.00 units

timebase

© Copyright IBM Corporation – AIX Collaboration Center 2006

 Processing Units Virtual Processors Shared Processor Pool
IBM System p
Specified in parts
of a processor

Capped

Partition 1 Virtual 5 2.5 2.5

0.5 1 Way
#1 Processor
10 milliseconds

Hypervisor Dispatch Cycle

Capped

Partition 2 Virtual
#2 0.5 2 Way Processors

3 5 3 5

10 milliseconds
Uncapped

Partition
1.2 4 Way 4 Virtual
#3 Processors

3 5 3 5

10 milliseconds

© Copyright IBM Corporation – AIX Collaboration Center 2006

Virtual Processors in Shared Processor Pool
IBM System p

• Virtual processors represent concurrent operating system operations

• Entitled Capacity (physical cpu) is spread across these virtual processors
• Optimal number of virtual processors depends on the workload
- Number of threads
- What threads are doing
• Number of virtual processors (Minimum and Desired) is obtained by:
- Rounding entitled capacity to next whole number
- Example
• Minimum = 0.50 (entitlement) -> 1 virtual processor minimum
• Desired = 2.25 (entitlement) -> 3 virtual processors desired
• Maximum number of virtual processors is 10x entitlement
- Do you want maximum 0.8 entitled over 8 virtual processors?
- Some art, experimentation warranted
- Some workloads need more concurrence, some need fewer and more
powerful virtual processors

© Copyright IBM Corporation – AIX Collaboration Center 2006

Understanding capacity allocation – An example
IBM System p

• A workload is run under different configurations.

• The size of the shared pool (number of physical processors) is fixed at
16.
• The capacity entitlement for the partition is fixed at 9.5.
• No other partitions are active.

© Copyright IBM Corporation – AIX Collaboration Center 2006

Uncapped – 16 virtual processors
IBM System p

Uncapped (16PPs/16VPs/9.5CE)

15

10

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Elapsed time

• 16 virtual processors.
• Uncapped.
• Can use all available resource.
• The workload requires 26 minutes to complete.

© Copyright IBM Corporation – AIX Collaboration Center 2006

Uncapped – 12 virtual processors
IBM System p

Uncapped (16PPs/12VPs/9.5CE)

15

10

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Elapsed time

• 12 virtual processors.
• Even though the partition is uncapped, it can only use 12 processing units.
• The workload now requires 27 minutes to complete.

© Copyright IBM Corporation – AIX Collaboration Center 2006

Capped
IBM System p

Capped (16PPs/12VPs/9.5E)

15

10

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Elapses time

• The partition is now capped and resource utilization is limited to the capacity entitlement of
9.5.
- Capping limits the amount of time each virtual processor is scheduled.
- The workload now requires 28 minutes to complete.

© Copyright IBM Corporation – AIX Collaboration Center 2006

Checkpoint
IBM System p

Match the following processor terms to the statements that describe them:

Dedicated Shared Capped Uncapped Virtual Logical

„ These processors cannot be used in Micro-partitions ________

„ Partitions marked as this may use excess processing capacity in the shared pool
________

„ There are two of these for each virtual processor if SMT is enabled ________

„ This type of processor must be configured in whole processor units ________

„ These processors are configured in processing units as small as one hundredth of a

processor ________

„ Partitions marked as this may use up to their entitled capacity but not more
________

„ These processor are dispatched in a time-sliced manner onto physical processors

by the POWER Hypervisor ________
© Copyright IBM Corporation – AIX Collaboration Center 2006
Summary
IBM System p

• This chapter covered

- Understand Micro-partitions
- Demonstrate how it was used to consolidate multiple applications
and increased server utilization

• References
- APV on System p5 (Redbook)
• APV on IBM System p5-sg247940.pdf
- APV on p5 Servers, Architecture and Performance (Redbook)
• APV Architecture and Performance-sg245768.pdf

© Copyright IBM Corporation – AIX Collaboration Center 2006