Université IBM i 2018

16 et 17 mai
IBM Client Center Paris

S48 – Best Practices for IBM i Memory Tuning for

Performance
Stacy L. Benfield
IBM i Performance Consultant - Lab Services Power Systems Delivery Practice
IBM i Large User Group (LUG) Program Manager
IBM, Rochester MN
stacylb@us.ibm.com
Agenda
 Memory Concepts and page faulting
 IBM i memory tuning features
 Tools used to identify memory issues
 Best practices for memory tuning

2
Importance of Memory
 The CPU(s) can’t do anything if the required data is not in memory
 Having to go to disk is slow, in relative terms

How to improve performance:

 Tuning can help your system use its memory better
 IBM i has technology built in to help manage memory automatically
 IBM i has excellent performance tools for analyzing memory
performance
 Adding hardware such as more memory or SSDs can reduce I/O wait
times
 Application changes can greatly reduce how often your jobs wait on disk
I/O

3
Bringing data into memory
 Implicit memory transfers – “page fault reads”
• Page faults – Synchronous memory transfer where jobs wait until I/O completes
• I/O Pending faults – waiting on your/or someone else’s I/O request to complete
• Collection Services & Job Watcher have a “wait bucket” to track this time

 Explicit memory transfers– “non-fault reads”

• Asynchronous memory transfer where OS pre-fetches data before application needs it
(jobs don’t have to wait)
– DB2 “read-ahead” or “asynchronous brings”
• SETOBJACC command puts an object into a memory pool
• Read part of CRTDUPOBJ and CPYF operations
• Collection Services & Job Watcher have a “wait bucket” to track this time

4
Types of I/O Requests for Reads and Writes
 Synchronous I/Os
• Processing waits until I/O completes
• Contributes to response/runtime
• Examples (reads):
SETOBJACC, CRTDUPOBJ, CPYF, Faults, …
• Asynchronous I/Os
• Processing concurrent with I/O
• Can turn synchronous
• Examples (reads):
DB2 asynchronous brings

5
Page Fault basics

The primary focus for memory analysis is to reduce the amount of time waiting on
disk faults.

 Page Faults are normal and expected

• It is the mechanism used to bring most things into memory
• Some things need to be brought into memory to be cleaned up / deleted

 A single page fault can bring multiple pages into memory

• Faults per second < pages per second
• We don’t typically care about pages per second

 Requested pages are placed into the memory pool the job is running in
• Pages can be shared between jobs
• A job can access a page in a different pool

6
Why do we care if page fault wait times are high
 High page fault wait times indicates pool tuning issues, unnecessary memory
usage, and/or lack of memory that will affect performance
– Longer interactive response times for I/O intensive transactions
– Longer batch run times
– Poor disk response times
– Less efficient query implementation methods

7
Page Faults – Faults Per Second vs. Wait Time - PDI
 Focus on time being spent waiting on faults, not the rate of faults
• Large memory pools can have a high number of faults, but little impact to wait times
 However, rates can be useful for monitoring and when you can’t get wait time easily
• Dashboard, System Monitors, WRKSYSSTS, etc.
• Also can indicate new workload has started up
 Which chart below is more helpful in determining if faulting might be a problem?
faults/sec wait time

8
Page Faults Wait Time at a Job Level
 Collection Services (and Job Watcher) collect disk page fault wait time both at a system
level, as well as an individual job/thread/task level
 ??? Is it more interesting to know that JOBX did 120 faults per second, or that it waited
370 seconds (out of 5 minutes) on disk page faults

9
IBM tasks associated with removing data from memory
System Controlled
 There are two tasks that are responsible for writing changed pages of memory
out to disk to make room for data being paged into memory.
• SMPOL001: low priority page out task
• SMPO0001: high priority page out task
– becomes active if the low priority task cannot keep up with demand
 High activity in these tasks, especially the high priority task, can indicate a need
for additional memory.

User Program Controlled

 Changed pages are written out when a job ends

10
Single-level Storage
 Unique to IBM i
 Main memory, traditional spinning disks, and solid-state drives all appear as one
address space
 Objects automatically managed by the OS, moving between memory and disk
 One copy of file/data can be shared by many users, jobs, programs
 Objects spread across drives for parallelism
 No need to create “table spaces”, “buffer pools”, etc. required by other platforms
 Think of memory as a giant cache for disk

11
IBM i Memory / Storage Hierarchy
HIGH LOW HIGH
L1 cache SPEED CAPACITY COST
Processor
Core/Chip L2 cache
L3 cache

“L4 Cache” (POWER8)

Memory

I/O Adapter Cache

Disk Buffer

Input / Output
(Disk)
LOW HIGH LOW
SPEED CAPACITY COST

12
POWER Memory / Storage Hierarchy
Size Speed / cycles
POWER8 L1 cache 32K instr + 64K data per core ~3 cycles
POWER8 L2 cache 512 KB per core ~10 cycles
POWER8 L3 cache 96 MB shared per chip ~30 cycles
POWER8 “L4 cache” 16 MB per memory chip* ~200 cycles
100x
Internal Memory 10s of GB per core ~300-800 cycles (~100ns)
Solid State Drives 100s of GB per drive <1 ms
10,000x
Hard Disk Drives Can be TBs per drive 1-5 ms

Range due to “distance” of access.

1 ms = 1000 µs = 1,000,000 ns
400 cycles is about 0.1 µs if 4.0 GHz

13
Memory Configuration – Verify there’s enough memory first
 A general rule of thumb for memory (minimum requirements) based off
internal benchmarks
• 32 GB/core for POWER8
• 24 GB/core for POWER7/7+
• 16 GB/core for POWER6
 Partitioning considerations
• If dynamically adding cores, also typically want to add memory (DLPAR)
• Uncapping can lead to an imbalance between CPU and memory
 Can utilize SSDs/Flash technology to improve page fault wait times

14
Memory Pools
 Memory pools are logical subdivisions of physical memory
 Used with subsystems to isolate memory usage by different applications
 Two types – shared and private

Shared Private
Subsystems single or multiple single

QPFRADJ yes no
Expert Cache yes no

15
Memory Pools

Machine Pool
User Pools
*INTERACT
*SPOOL
*SHRPOOL1-n
*BASE

(residual)

16
Pool Maximum Activity Level
 The maximum number of threads in the pool that can use the CPU(s)
concurrently
• Threads without an activity level are ineligible to run
 Does not apply to the machine pool
• No jobs run in the machine pool
 Can be adjusted + or - by the Performance Adjustor
• Adjuster conservative on decreasing
 Considerations
• Generally want high enough to avoid transitions to ineligible
– There is an “Ineligible Waits time” bucket in CS and JW
• Setting too low can lead to severe performance problems
• Setting too high can lead to more faulting

17
IBM i memory tuning features
 Expert Cache
 Automatic performance adjust system value (QPFRADJ)
 SETOBJACC command
 DB2 keep in memory

18
Expert Cache (*CALC)
 Paging parameter for shared storage pools
• *FIXED
• *CALC – enables expert cache
 What it does:
• Monitors the I/O reference pattern for database files
• Reduces I/O operations by adjusting the size and type of I/Os
 Cannot be used by:
• Private pools (can be used by running an API – QWCCHGTN)
• Machine pool
 Activity you may see:
• SMXCSPRVSR (Expert Cache supervisor) task
• SMXCAGERnn tasks (1 per *CALC pool, 01 => pool 2)
 Expert Cache almost always provides benefit and should be enabled

19
Expert Cache Enablement

20
The Performance Adjuster
 Enabled using the QPFRADJ system value.
 Will manage the size of the shared memory pools for you.
• Will also adjust the maximum activity level.
 Uses complex algorithms to ensure your pools are operating at peak
efficiency.
 Refer to The Performance Adjuster (QPFRADJ) experience report on the IBM
i Information Center.
http://publib.boulder.ibm.com/infocenter/iseries/v7r1m0/index.jsp?topic=%2Fexperience%2Fwork3abstract.htm

21
Automatic Changes in Pool Sizes - PDI

22
Turning QPFRADJ On or Off
The Performance Adjuster is controlled by the System Value QPFRADJ
 0=No adjustment User managed!

 1=At IPL only based on static information

 2=At IPL and automatic (default)
 3=Automatic only Recommended

23
Tuning QPFRADJ (WRKSHRPOOL)

24
Tuning QPFRADJ

WRKSHRPOOL:

25
Tuning shared pools when QPFRADJ is ON
 Use WRKSHRPOOL command to tune shared memory pools
 Determine min/max size for critical pools
• Monitor max active settings
 Set a range of pool priorities.
 Java/WebSphere pools minimum size should always be enough to contain all JVMs in the pool
 Set max on memory intensive pools to limit impact to other jobs on the system
 Large changes to the size of the memory pool can cause the query optimizer to rebuild access plans
which can contribute to poor performance
 If see pools sitting at min size a lot, consider decreasing its min size
 If see pools hitting max size, consider increasing max size.
 Keep total of minimum sizes < 70% of memory to allow QPFRADJ some flexibility
 When adding memory, may need to adjust min/max
• When memory is for a particular workload, rather than a general upgrade
• Values are percentages of total

Refer to The Performance Adjuster (QPFRADJ) experience report on the IBM i Information Center.
http://publib.boulder.ibm.com/infocenter/iseries/v7r1m0/index.jsp?topic=%2Fexperience%2Fwork3abstract.htm
26
SETOBJACC to ‘pin’ objects in memory
 Allows selected objects (database file, index or program) to be "pinned" in memory to
reduce I/O
• Typically, define a private pool where no jobs run
• Can use a shared pool. Set Min/Max to keep QPFRADJ from shrinking
 May increase memory requirements
 Protects objects from “demand paging” activity
 Objects loaded into memory very quickly by a single thread
 Typically used to improve performance of batch jobs
 Run a CLRPOOL command before loading objects in pool
 Re-run SETOBJACC periodically to pin changed pages (updates/inserts)
 Can load and purge objects programmatically during batch processing

SETOBJACC/CLRPOOL Command Technote:

http://www.ibm.com/support/docview.wss?uid=nas1dc0a2297bdaefddb86256d6c0069907f
27
DB2 KEEPINMEM to ‘pin’ objects in memory
 New in IBM i 7.1
 Can be used with tables and indexes
 Objects are brought into memory when first accessed by SQE queries
 Brought in asynchronously and using parallel I/O
(vs. SETOBJACC which uses 1 thread and synch IO)
 Usage:
• CHGPF FILE(library/table) KEEPINMEM(*YES)
• CHGLF FILE(library/index) KEEPINMEM(*YES)
 The MEMORY_POOL_PREFERENCE parameter in the QAQQINI file determines
which pool the objects will be held in unless the object is already in memory
 In IBM i 7.2. the KEEP IN MEMORY memory-preference support has been
extended to the DDL statements: ALTER TABLE, CREATE INDEX, CREATE TABLE
and DECLARE GLOBAL TEMPORARY TABLE

28
Memory Pool and Subsystem Creation Recommendations
 Minimize user work in *BASE (in general)
 Put disparate workloads into separate pools
 Often want to run SQL in its own pool
 Heavy memory use when not critical
• Often see software replication jobs in own pool with lots of faulting, but still keeping up
 Java / WebSphere should run in its own memory pool
• These environments cannot tolerate faulting
 Controlling potentially inefficient work
• QZDA jobs in own pool by IP for controlling ad-hoc queries
 Max active settings
• If not using QPFRADJ set max active high enough to avoid transitions to ineligible
• If running 6.1 don’t set too high for pools running SQL
 Settings related to QPFRADJ (min/max size, etc.)
 If see high priority page out task (SMPO0001), pool needs more memory.

29
Tools Used for Memory Analysis
 Real Time Tools
• Green screen commands / IBM Navigator for i
– WRKSYSSTS, WRKACTJOB
• IBM Navigator for i Dashboard
• IBM Navigator for i Monitors (New in IBM i 7.2)
 Collector based
• Collection Services – System and Job level memory related wait times, fault rates, and
pool sizes
• Job Watcher – Object waited on, Call Stacks, and SQL statement
 GUI Tools
• Performance Data Investigator (“PDI” Web based, part of IBM Navigator for i) –
Collection Services, Job Watcher, Disk Watcher, limited PEX
• IBM iDoctor for i (Windows based) – Collection Services, Job Watcher, Disk Watcher,
PEX Analyzer
• System i Navigator

30
WRKSYSSTS – Faulting by Memory pool
 Can be invoked from green screen or web based IBM navigator
for i
 Faults per second column doesn’t show real fault wait times
 Monitor the max active and ineligibles

31
Memory Pools (WRKSYSSTS)
GUI advantage

or

32
WRKACTJOB – Page faulting by job
The Page fault rate column is available of the IBM Navigator for i interface

33
Navigator Dashboard

Also available in iAccess Mobile: http://system.name:2001/iamobile 34

Navigator for i System Monitors (7.2)

35
IBM Graphical Analysis Tools
IBM provides two powerful tools to aid in making your analysis more efficient and
productive:

 Performance Data Investigator

 IBM iDoctor for IBM i

Both solutions support data analysis (varying degrees) for the 4 collectors:
• Collection Services
• Job Watcher
• Disk Watcher
• Performance Explorer (PEX)

36
Performance Data Investigator (PDI)
 Browser (web) based solution
 Integrated as part of IBM i OS
 Included in IBM Navigator for i

37
IBM iDoctor for IBM i
 Microsoft Windows based client
 Service/Support offering
 Deep Job Watcher and PEX analysis capability

38
Graphical Analysis Tools
 You have two graphical interfaces for performance data analysis…
– Which is right for you?
Feature iDoctor PDI
Interface Windows client Browser
Wait Analysis Yes Yes
Collection Services Yes Yes
Job Watcher Yes (In-depth) Yes
Disk Watcher Yes Yes
Performance Explorer Yes (In-depth) Profile collections only
Level of analysis provided Deep Basic to Medium
Job Watcher Monitors (Built –in) Yes No
User Defined graphs and queries Yes Yes
Update Frequency Quarterly Twice Yearly

Support Email idoctor@us.ibm.com Standard SWMA

Chargeable Yearly license for each component (by Collection Services & Health Indicators at no additional charge with i
serial number)
Disk Watcher, Database, and Performance Explorer included with base PT1
Job Watcher (Performance Tools LPP) product – Option 1 Manager feature
–Includes Job Watcher, Collection
Job Watcher is an additional option of PT1 and has an additional charge -
Services Investigator, and Disk
Option 3 Job Watcher
Watcher
PEX Analyzer
DS8K graphs & VIOS Investigator Yes No
Multinational language support No Yes
39
Collection Services - Memory Related Data
 What will be analyzed using CS?
• Page fault wait time for system, job, thread, memory pool etc.
– At a job level, can be useful to know if the wait time occurred primarily during
start-up time, or consistently while active
• Rate of faulting in each memory pool
• Memory pool size/config of the jobs with high page faults

40
Memory Analysis - CSI

 The first step is to identify page fault wait times in a collection services member

41
Memory Analysis - CSI
 A good place to start is the Collection overview time signature.
 We want to see the relative amount of time being spent waiting on disk faults.

42
Memory Analysis - CSI
 Identify intervals with high page fault wait times (tan)
 Wait times could be skewed some by large collection services intervals

43
Memory Analysis – Memory Pool Graphs in CSI
 No change in pool size indicates that performance adjuster system value (QPFRADJ)
was not on during this collection

44
Memory Analysis - CSI
 Highlight intervals with high page fault waits and right click to rank by memory pool

45
Identify top jobs in the pool with most page faults

46
Worst page faulting jobs for all pools
 The following chart shows worst offending jobs and the pools they run in
 We added the Pool No. to the X-Axis label.

47
Memory Analysis - CSI
 Right click on jobs of interest to get wait time signature
 It is often useful to group threads in different ways such as by generic name, current user id etc.

48
Memory Analysis - CSI
 Click on any interval to get the specific wait statistics

49
Memory Analysis - CSI
 The wait time details are shown in the tabs below

50 50
Collection Services Memory analysis summary
 We can tell how much faulting is happening at a system level
• And at a job or thread or task level
 We can tell what pools have the most faulting
 We can see if memory pools are changing in size
 We can’t tell what the jobs are doing that is causing faulting
• Or what they are faulting on

 For additional details….

• Job Watcher data can provide answers to many questions.

51
Job Watcher Memory Related Data
 Job Watcher has similar information as Collections Services. Key
differences are:
• Intervals typically much shorter (5-10 seconds vs. 5-15 minutes)
• Additional data is collected
– Call stacks are collected
– Objects being waited on are collected
– SQL being run is collected

 Things are different once we start looking at interval details

52
Launching Job Watcher
• Select Job Watcher and Click Launch

53
Memory Analysis - Job Watcher
 Entry Point is the same as CSI – Collection Overview time Signature, but the chart will show shorter intervals
for a more obvious impact of all wait times

54
Job Watcher drill down for memory analysis
 The same drill down process can be used here as it was in CSI

55
Analyzing job-threads with high page faulting
 Threads ranked by longest page fault wait times

56
Investigating wait signature over time
 Use your mouse to fly over the intervals to see the CPU Utilization and page fault wait times in seconds

57
Reviewing Call stacks
 Right click on the intervals of interest and click on “Display call stack”

58
Call Stack information
 The call stack will show the object faulted on, pool, and program driving it

Very useful!!!

59
Example: Finding cause for intermittent page faults in a job
 In this example, a job shows intermittent faults

60
Identify program causing page faults
 The user program here is calling an SQL statement that then calls the QDBGETMQO (table or index scan) MI
program

61
Identify the SQL statement causing high page faults
 The SQL tab will often show which statement was run to cause the faulting

62
Job Watcher Memory Analysis Summary

 We can tell with more granularity what CS data showed us:

• The job waiting the most on disk faults
• The type of faults (DB vs. non-DB)
• The average wait time on faults
• The pool the job is running in
• Etc.
 Additionally, we now know:
• The object being faulted on
• The program running
• The OS operation causing the faulting
• May also get SQL statement and host variables if running SQL

63
Navigator SQL Plan Cache Snapshot Analysis
 Run a plan cache snapshot against the tables being faulted in the most

64
Plan Cache Snapshot Analysis
 Filter the plan cache snapshot by objects

65
Data Management Best Practices for Memory

 Reduce file sizes

 Implement an index strategy
 Review index page size
 Remove deleted records (RGZPFM command)
 Purge historical data
 Consider adding SSD arms for those objects that have to be faulted in
 Avoid making any pool with activity too small – the faulting caused
additional I/O, which in turn can affect response times for other jobs

66
IBM i Memory Tuning Best Practices - Summary
 Verify the memory configuration
• Memory per processor
• Processor affinity score
• Reduce IO time by removing I/O bottlenecks
 Verify the pools are tuned correctly
• Min/max pool size set up
• Jobs running in the right pool
• IBM i performance features are being used
 Tune application data access
 Optimize SQL queries
 Verify files size are right for an OLTP environment
• Remove deleted records
• Purge historical data

67
Questions?

68
References

69
 IBM i Performance FAQ a MUST read!
October 2017 update (watch for a Spring 2018 soon!):
https://www-01.ibm.com/common/ssi/cgi-bin/ssialias?htmlfid=POW03102USEN

70
IBM i Web Sites with Performance Information
 IBM Knowledge Center:
• 7.2 Performance
• 7.3 Performance

 IBM i Performance Management:

i Performance Management

 developerWorks:
‒ IBM i Performance Tools: developerWorks Performance Tools
‒ IBM i Performance Data Investigator: developerWorks PDI

 IBM iDoctor for IBM i: iDoctor

 IBM i Wait Accounting information:

‒ Job Waits Whitepaper
‒ KnowledgeCenter: The basics of Wait Accounting
‒ developerWorks: IBM i Wait Accounting

71
 A Redbooks publication!

http://www.redbooks.ibm.com/redbooks/pdfs/sg247808.pdf

72
IBM i 7.2 Technology Refresh Updates

Covers the 7.2 content through

Technology Refresh 1

Section 2.8 – Performance

Section 8.6.7 – Job level SQL stats in

Collection Services

73
 IBM i Performance Analysis Workshop
Learn the science and art of performance analysis, methodology and problem solving
Managing and analyzing the data can be quite complex. During this workshop, the IBM Systems Lab Services IBM i team will share useful techniques for analyzing performance data on key
IBM i resources, and will cover strategies for solving performance problems. It will aid in building a future foundation of performance methodology you can apply in your environment.
Overview:
• Topics covered include:
– Key performance analysis concepts
– Performance tools
– Performance data collectors (Collection Services, Job Watcher,
Disk Watcher, and Performance Explorer)
– Wait accounting
• Core methodology and analysis of:
– Locks
– Memory
– I/O subsystem
– CPU
• Concept reinforcement through case studies and lab exercises
• Discussions on theory, problem solving, prevention and best practices
Workshop details:
•Intermediate IBM i skill level
• 3-4 day workshop, public or private (on-site)
– For general public workshop availability and enrollment:
IBM i Performance Analysis Workshop
– For public workshop availability and enrollment in France, please contact Philippe Bourgeois at
pbourgeois@fr.ibm.com or Françoise Laurens at f_laurens@fr.ibm.com
• For additional information, including private workshops, please contact Eric Barsness
at ericbar@us.ibm.com or Stacy Benfield at stacylb@us.ibm.com, members of Systems Lab Services
IBM Systems Lab Services Power Systems Delivery Practice - ibm.com/systems/services/labservices - ibmsls@us.ibm.com
© 2018 IBM Corporation
 IBM i Performance and Optimization Services

The IBM i Performance and Optimization team specializes in resolving a wide variety of performance problems. Our team of
experts can help you tune your partition and applications, including:
• Reducing batch processing times
• Resolving SQL query and native IO performance problems
• Tuning RPG, COBOL, C, and Java (including WebSphere Application Server) programs
• Removing bottlenecks, resolving intermittent issues
• Resolving memory leaks, temporary storage growth problems, etc.
• Tuning memory pools, disk subsystems, system values, and LPAR settings for best performance
• Optimizing Solid State Drive (SSD) performance
• Tuning client interfaces such as ODBC, JDBC, .Net and more

Skills transfer and training for performance tools and analysis also available!
Contact Eric Barsness at ericbar@us.ibm.com for more details.
www.ibm.com/systems/services/labservices
75
197
And finally…..
 Thank you

Don't forget to fill-in the

feedback form!
www.ibm.com/power/i

78
 Special notices
This document was developed for IBM offerings in the United States as of the date of publication. IBM may not make these offerings available in other countries,
and the information is subject to change without notice. Consult your local IBM business contact for information on the IBM offerings available in your area.
Information in this document concerning non-IBM products was obtained from the suppliers of these products or other public sources. Questions on the
capabilities of non-IBM products should be addressed to the suppliers of those products.
IBM may have patents or pending patent applications covering subject matter in this document. The furnishing of this document does not give you any license to
these patents. Send license inquires, in writing, to IBM Director of Licensing, IBM Corporation, New Castle Drive, Armonk, NY 10504-1785 USA.
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.
The information contained in this document has not been submitted to any formal IBM test and is provided "AS IS" with no warranties or guarantees either
expressed or implied.
All examples cited or described in this document are presented as illustrations of the manner in which some IBM products can be used and the results that may
be achieved. Actual environmental costs and performance characteristics will vary depending on individual client configurations and conditions.
IBM Global Financing offerings are provided through IBM Credit Corporation in the United States and other IBM subsidiaries and divisions worldwide to qualified
commercial and government clients. Rates are based on a client's credit rating, financing terms, offering type, equipment type and options, and may vary by
country. Other restrictions may apply. Rates and offerings are subject to change, extension or withdrawal without notice.
IBM is not responsible for printing errors in this document that result in pricing or information inaccuracies.
All prices shown are IBM's United States suggested list prices and are subject to change without notice; reseller prices may vary.
IBM hardware products are manufactured from new parts, or new and serviceable used parts. Regardless, our warranty terms apply.
Any performance data contained in this document was determined in a controlled environment. Actual results may vary significantly and are dependent on many
factors including system hardware configuration and software design and configuration. Some measurements quoted in this document may have been made on
development-level systems. There is no guarantee these measurements will be the same on generally-available systems. Some measurements quoted in this
document may have been estimated through extrapolation. Users of this document should verify the applicable data for their specific environment.
 Special notices (cont.)
IBM, the IBM logo, ibm.com AIX, AIX (logo), AIX 6 (logo), AS/400, BladeCenter, Blue Gene, ClusterProven, DB2, ESCON, i5/OS, i5/OS (logo), IBM Business Partner (logo),
IntelliStation, LoadLeveler, Lotus, Lotus Notes, Notes, Operating System/400, OS/400, PartnerLink, PartnerWorld, PowerPC, pSeries, Rational, RISC System/6000, RS/6000,
THINK, Tivoli, Tivoli (logo), Tivoli Management Environment, WebSphere, xSeries, z/OS, zSeries, AIX 5L, Chiphopper, Chipkill, Cloudscape, DB2 Universal Database, DS4000,
DS6000, DS8000, EnergyScale, Enterprise Workload Manager, General Purpose File System, , GPFS, HACMP, HACMP/6000, HASM, IBM Systems Director Active Energy
Manager, iSeries, Micro-Partitioning, POWER, PowerExecutive, PowerVM, PowerVM (logo), PowerHA, Power Architecture, Power Everywhere, Power Family, POWER Hypervisor,
Power Systems, Power Systems (logo), Power Systems Software, Power Systems Software (logo), POWER2, POWER3, POWER4, POWER4+, POWER5, POWER5+, POWER6,
POWER6+, System i, System p, System p5, System Storage, System z, Tivoli Enterprise, TME 10, Workload Partitions Manager and X-Architecture are trademarks or registered
trademarks of International Business Machines Corporation in the United States, other countries, or both. If these and other IBM trademarked terms are marked on their first
occurrence in this information with a trademark symbol (® or ™), these symbols indicate U.S. registered or common law trademarks owned by IBM at the time this information was
published. Such trademarks may also be registered or common law trademarks in other countries. A current list of IBM trademarks is available on the Web at "Copyright and
trademark information" at www.ibm.com/legal/copytrade.shtml

The Power Architecture and Power.org wordmarks and the Power and Power.org logos and related marks are trademarks and service marks licensed by Power.org.
UNIX is a registered trademark of The Open Group in the United States, other countries or both.
Linux is a registered trademark of Linus Torvalds in the United States, other countries or both.
Microsoft, Windows and the Windows logo are registered trademarks of Microsoft Corporation in the United States, other countries or both.
Intel, Itanium, Pentium are registered trademarks and Xeon is a trademark of Intel Corporation or its subsidiaries in the United States, other countries or both.
AMD Opteron is a trademark of Advanced Micro Devices, Inc.
Java and all Java-based trademarks and logos are trademarks of Sun Microsystems, Inc. in the United States, other countries or both.
TPC-C and TPC-H are trademarks of the Transaction Performance Processing Council (TPPC).
SPECint, SPECfp, SPECjbb, SPECweb, SPECjAppServer, SPEC OMP, SPECviewperf, SPECapc, SPEChpc, SPECjvm, SPECmail, SPECimap and SPECsfs are trademarks of the
Standard Performance Evaluation Corp (SPEC).
NetBench is a registered trademark of Ziff Davis Media in the United States, other countries or both.
AltiVec is a trademark of Freescale Semiconductor, Inc.
Cell Broadband Engine is a trademark of Sony Computer Entertainment Inc.
InfiniBand, InfiniBand Trade Association and the InfiniBand design marks are trademarks and/or service marks of the InfiniBand Trade Association.
Other company, product and service names may be trademarks or service marks of others.
End of Presentation material…..