for SAP Solutions on

cavalcadejewelSoftware and s/w Development

Nov 18, 2013 (3 years and 9 months ago)

405 views

©
2012
IBM Corporation

Place photo here

Matthias
Koechl

February 2012

Update on
Virtualization
C
apabilities
for
SAP
S
olutions
on

IBM Power Systems™

©
2012
IBM Corporation

Content


POWER7 Processor Characteristics


SMT Modes


Entitlement versus virtual CPUs


PowerVM Advisors



Memory Virtualization


Advanced Memory Expansion (AME
)


Advanced Memory Sharing (AMS
)



PowerVM Live
Partition Mobility



AIX Workload
Partitions (WPARs)



SAP
and PowerVM
Integration


POWER metrics in
CCMS


SAP Cloud Computing



2

ISICC Walldorf, MK

©
2012
IBM Corporation

Content


POWER7 Processor Characteristics


SMT Modes


Entitlement versus virtual CPUs


PowerVM Advisors



Memory Virtualization


Advanced Memory Expansion (AME
)


Advanced Memory Sharing (AMS
)



PowerVM Live
Partition Mobility



AIX Workload
Partitions (WPARs)



SAP
and PowerVM
Integration


POWER metrics in
CCMS


SAP Cloud Computing



3

ISICC Walldorf, MK

©
2012
IBM Corporation

4

Leading SAP Benchmarks for Power Systems

Source:
http://www.sap.com/benchmark/


Results listed with “processors” / cores / threads.


126,063

70,032

37,000

32,000

25,500

25,160

24,910

24,003

24,000

492

547

578

125

318

314

311

300

300

0
100
200
300
400
500
600
0
20,000
40,000
60,000
80,000
100,000
120,000
Users per Core

SD Benchmark Users

Fujitsu

P’Quest

1800E2

8/80/160

SQL Server

#2011017

IBM

Power 795

32/256/1024

DB2

#2010046

HP

DL980

8/80/160

SQL Server

#2011021

IBM

Power 795

16/128/512

DB2

#2010042

SAP SD benchmark Users

Users per CPU

All results are as of 12/1/11.

Sun/Fujitsu

M9000

64/256/512

Oracle

#2009046

NEC

Express5800

8/80/160

SQL Server

#2011023

IBM

Power 780

8/64/256

DB2

#2010013

IBM

X3850 X5

8/80/160

DB2

#2011034


HP

DL980

8/80/160

SQL Server

#2011020


SAP SD Benchmarks are a spezialied workload patterns with parallel transaction workload.


They provide a good proof
-
point for system capacity sizing, not transaction response time


They exploit POWER7 SMT4 capabilities and system bandwidth.


Customer workload has different behaviour

©
2012
IBM Corporation

POWER7 based Servers and Application performance


By more Cores and more parallel Threads (SMT4), POWER7 based servers
deliver excellent „SAPS“ values.


By nature of the benchmark SAPS are a good measure for
system throughput
.


Performance of an isolated user transaction or batch program is dominated by
single thread performance
.
These do NOT benefit from SMT effect
.



Whenever estimating single thread dominated SAP workload performance, avoid
system sizing purely based on SAPS values.


In particular important when doing an upgrade sizing from existing POWER platforms with lower SMT gains
(see left diagram)
:



5

ISICC
Walldorf
, MK


Avoid running POWER7 systems in POWER6 Mode


only SMT2 will be exploited.


©
2012
IBM Corporation

6

POWER7 Multi
-
threading Options

0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
SMT1
SMT2
SMT4


Certain SAP workload characteristics
benefit from different SMT modes


Always differentiate “system capacity” =
SAPS and “performance” of a single thread =
response time




SMT1:

Largest unit of execution work



SMT2:

Smaller unit of work, but provides greater
amount of execution work per cycle



SMT4:

Smallest unit of work, but provides the
maximum amount of execution work per
cycle

SAP Batch

SAP Dialog

Tx A

Tx B

Tx C

Tx D

Tx A

Tx B

Sequential Job A

Single
-
threaded

Aggregate throughput/ POWER7 core

ISICC
Walldorf
, MK

©
2012
IBM Corporation

POWER7 Virtual
Processors &
Entitlement (1)

7

ISICC Walldorf, MK


Entitlements specify guaranteed CPU resources for a LPAR



Too low an entitlement means a partition’s portion of a dispatch cycle is small, and work may be spread across
many dispatch cycles instead of fewer.



This artificially introduces relative hypervisor overhead compared to SAP workload



Cache hit ratio may be negatively impacted



Virtual Processors (
vCPUs
) specify the maximum of accessible CPU resources for a
LPAR


Too
many Virtual Processors being dispatched in
a constrained
environment reduces
performance.



tests showed up to
-
30% performance impact for wrongly designed LPARs (entitlement /
vCPU

mismatch)
running concurrent DB workloads, compared to best
-
practices LPAR setup


Processor Folding mitigates the
overhead
of
spreading the partition over many more
vCPUs
, but
does not
completely eliminate
it
(
-
20% vs.
-
30
%).


Is sub
-
optimal parameterization
rather than PowerVM inefficiency



S
preading
a
LPAR
over more
vCPUs

than
needed increases the likelihood of lock
contention
during LPAR scheduling


Databases
are particularly lock intensive, so conservative usage of
vCPUs

is beneficial



©
2012
IBM Corporation


Entitlement settings should be realistic


Size LPAR entitlement in a way that it covers average high load phases


Ideal entitlement requires use of few incremental
vCPUs

only to cover absolute peak load



Peak loads > 100% of entitlement are not critical in shared processor pools


Adjust entitlement according to monitoring





Virtual Processor (
vCPU
) settings


Although it provides most flexibility in cross LPAR peak load compensation, avoid too large count of
vCPUs

beyond the
cpu

capacity allocated by a LPAR’s entitlement


If there are always free pool cycles,
vCPU

setting is less critical


Use
(
topas


C or
lparstat

“app
”) to check


If the pool is constrained, however, the optimal performance will always be by keeping the
vCPU

to
physical core ratio as close as possible



Read “POWER7 Virtualization
Best
Practice
Guide”
https://
www.ibm.com/developerworks/wikis/display/WikiPtype/Performance+Monitoring+Documentation


Understand the differences in performance reporting across the
POWER generations
http://www.ibm.com/developerworks/wikis/display/WikiPtype/Understanding+Processor+Utilization+on+POWER+Systems+
-
+
AIX





8

ISICC Walldorf, MK

POWER7 Virtual Processors & Entitlement
(2)

©
2012
IBM Corporation

PowerVM Performance Advisors


SAP relevant Advisors can be downloaded from IBM Developer Works:


Virtualization Performance LPAR Advisor


https://www.ibm.com/developerworks/wikis/display/WikiPtype/PowerVM+Virtualization+
performance+lpar+advisor


VIOS
Advisor


https://
www.ibm.com/developerworks/wikis/display/WikiPtype/VIOS+Advisor

9

ISICC Walldorf, MK

Power System

Power System

PC

FTP

©
2012
IBM Corporation

Sample Outputs from PowerVM Advisors

10

ISICC
-

MK, 02/2012


Contains sections for detailled
environment reporting and checks for
current settings


OK

Can be improved

VIO

LPAR

10

©
2012
IBM Corporation

Content


POWER7 Processor Characteristics


SMT Modes


Entitlement versus virtual CPUs


PowerVM Advisors



Memory Virtualization


Advanced Memory Expansion (AME
)


Advanced Memory Sharing (AMS
)



PowerVM Live
Partition Mobility



AIX Workload
Partitions (WPARs)



SAP
and PowerVM
Integration


POWER metrics in
CCMS


SAP Cloud Computing



11

ISICC Walldorf, MK

©
2012
IBM Corporation

Active Memory™ Expansion (AME)


Innovative POWER7 technology


For AIX 6.1 or later


For POWER7 servers


Uses compression/decompression to
effectively expand the true physical memory
available for client workloads


Often a small amount of processor resource
provides a significant increase in the effective
memory maximum



AME has been tested with SAP ERP
workload, and delivered excellent expansion
factors.


A detailed White Paper including SAP
ERP Measurements can be downloaded
from:

ftp://ftp.software.ibm.com/common/ssi/sa/wh/n/pow03038usen/PO
W03038USEN.PDF


Fully supported for SAP production
environments by
DB2 V9 and Oracle 11g ff.


Watch
SAP
-
Note 1464605

“POWER7 Active
Memory Expansion (AME)”

for updates and
more details about AME




12

ISICC Walldorf, MK

Power Hypervisor

Hardware

Management

Console

New LPAR Attributes


Active Memory
Expansion On/Off


Memory Expansion
Factor

POWER7 System

LPARs’
memory

compressed memory

uncompressed mem.

©
2012
IBM Corporation

13

ISICC Walldorf, MK

AME Business value

AME

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

LPAR

System Physical
Memory

LPAR

LPAR

LPAR

LPAR

….


Enable more LPAR’s per server


Active memory expansion reduces the physical
memory requirements of existing LPAR’s


Existing LPAR’s physical memory sizes can be
reduced


Free memory capacity can be used to create
more LPAR’s


Supports SAP consolidation scenarios





Increase a LPAR’s effective memory capacity


Active memory expansion can increase the
effective memory capacity of a LPAR


Enabling active memory expansion for a LPAR
and keeping the LPAR’s physical memory size
unchanged increases the memory available to a
workload


Supports natural growth

AME

LPAR’s Effective
Memory Capacity

Physical Memory

Physical Memory

LPAR’s Effective
Memory Capacity

Expanded
Memory Capacity

System Physical
Memory

©
2012
IBM Corporation

Sample SAP ERP Workload test results at constant throughput

SAP 2tier setup within a single Partition (DB
-

and AppServer)


Test configurations held total memory constant at 14.25 GB, varying mix of real and gained
memory.


Number of cores was kept constant at 4 cores.


Individual load results will vary depending on expansion factor of the SAP applications and
data and available CPU resource




50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
100%
0%
20%
40%
60%
80%
100%
120%
140%
160%
180%
CPU Utilization

Relative Memory Expansion

AME Memory Gain SAP ERP

Memory gain by AME
CPU Utilization
14

ISICC Walldorf, MK

75% more memory for
1% additional CPU

111% more memory for

15% additional CPU

©
2012
IBM Corporation

AME Tests with complex SAP Retail DB2 ETL Load


Using a Power 750 w/ AME and Power 570, the following PoC has been performed:


Load 2 years of POS data of a large retail company into SAP NetWeaver BW 7.3 through the layers
of the Layered Scalable Architecture for POS data


~20 billion records stored in the InfoProviders of the Layered Scalable Architecture for POS data
management


Database size = 6.4 TByte


15

ISICC Walldorf, MK

©
2012
IBM Corporation

Retail ETL AME Test results 1


BI Application Server


AME Expansion > 2x with minor impact on both CPU utilization and job runtime
.



Measurement with AME off @ 20GB phys. showed performance degradation caused by
OS paging, while AME on @ 16GB phys. worked well

16

ISICC Walldorf, MK

AME off









Less CPU util

because of

wait I/Os


caused by


Paging


©
2012
IBM Corporation

Retail ETL AME Test results 2


BI DB
-
Server


DB2 Compression was turned ON during all
tests



AME Expansion worked for DB2 Server instance


Less expansion usable factor,
approx

1,5x


Not linked to DB2 compression, consistent with non
-
compressed DB observations


17

ISICC Walldorf, MK

©
2012
IBM Corporation

AME for a Memory
-
Intensive SAP SCM Component (
liveCache
)

18

ISICC Walldorf, MK

DB2


Virtual I/O Server

DB2 LUW 9.7

w/ compression

OS &
Pagespace

liveCache
persistence

liveCache

SAP SCM APO Application Server

2
-
tier SCM
-

POWER7 LPAR: 4 cores

Buffer Pool

“In
-
Memory MaxDB”

Σ
=
15.6 GB

10 GB

Physical
Memory = 8.5
GB

3,5 GB

1.5 GB

0,6 GB



System was tested with three Supply Chain Management (SCM) planning scenarios


Supply Network Planning, Demand Planning, Production Planning/Detailed Scheduling


White Paper available at
TechDocs

http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/WP102031


to accommodate different memory access patterns


liveCache

fill rate was >70%


Interestingly, this very memory intensive workload is well behaving with AME
!



AME=ON

Expansion

Factor

~2x

©
2012
IBM Corporation

AME effect at OS Level and Impact on SCM Throughput

19

ISICC Walldorf, MK

SNP2
1.5
1.5
1.8
2
2.6
2.6
1 416 309
1 344 789
1 351 351
1 348 315
1 317 716
1 286 174
0
0.5
1
1.5
2
2.5
3
30_1,5_ipl
35_1,5
31_1,8
32_2,0
33_2,6
34_2,6
AME factor
1 000 000
1 050 000
1 100 000
1 150 000
1 200 000
1 250 000
1 300 000
1 350 000
1 400 000
1 450 000
SNP orders/hour
AME factor
throughput
68% DIMM
savings

AME Exp. = 1,5 IPL 1,5 1,8

2,0
2,6 2,6 reptd.

Paging Activity under Load
0
200
400
600
800
1 000
1 200
1 400
1 600
1 800
2 000
30_1,5_ipl
35_1,5
31_1,8
32_2,0
33_2,6
34_2,6
OS Pgs/Sec
0
200
400
600
800
1 000
1 200
1 400
1 600
1 800
2 000
Compression
Pgs/Sec
pgsin
pgsout
compressed pgins
compressed pgouts
OS Paging

AME
compression is
no more sufficient,
OS paging required

AME Exp. = 1,5 IPL 1,5

1,8
2,0 2,6 2,6 reptd.

OS (AIX)

AME

©
2012
IBM Corporation

AME Sizing Hints


Ideally, use the
amepat

tool (part of AIX 6ff)

and have it run against existing workload.


Recommends individual settings for

AME expansion factor


Factor can be changed dynamically,


once AME has been enabled








Customer
feedback
from
live SAP environments
and from
internal
test shown
before.
As initial recommendations for AME settings apply these factors:



AME
Expansion Factor SAP Application
-
Server ABAP = max. 2,0


AME Expansion Factor SAP Application
-
Server JAVA = ca. 1,1


AME Expansion Factor SAP DB
-
Server = max. 1,6


20

ISICC Walldorf, MK

©
2012
IBM Corporation

21

ISICC Walldorf, MK

POWER Active Memory Sharing (AMS)

Virtualization Control Point (VCP)

SM

Partition 1



Paging VIOS (PSP)

(1 GB)

Page In / Out

V
A
S

I

Virtualization Control Point (VCP)

Shared Memory
Pool

(16 GB)

Free Memory
Pool

(5.5 GB)

Hypervisor
Memory


(1.5 GB)

Dedicated
Memory

(9 GB)

Physical Memory (32 GB)

F
C

SM Partition
2


SM
Partition 3


Dedicated
Memory
Partition 4
(8 GB)

Paging Devices

vSCSI
Server

CMM

CMM

CMM

Shared Memory Manager (SMM)

Hypervisor

Page
Donation

Shared Memory Pool
:

16 GB

Paging VIOS

Paging Devices

SM P1:
Desired Mem= 12GB

I/O Entitled Mem: 100 MB

SM P2:
Desired Mem= 8GB

I/O Entitled Mem: 100 MB

SM P3:
Desiredt Mem= 4GB

I/O Entitled Mem: 100 MB


POWER6™ feature


Administration comparable to
SPLPARs



No instantaneous memory allocation
as with
cpu

cycles!


“Ballooning” policies for memory
loaning


Not formally supported by SAP for
production use.


Not yet thoroughly tested with SAP
applications.

-

PoCs

with small SAP instances work
fine

-

Scalability for large memory sizes ?


Formal DB vendor support statements
pending
(DB2, Oracle,
MaxDB
)



Pilot program
(non
-
production environments)

started in 11/09, acquiring additional
pilot customers.


Technical
Paper on PW:

http://www.ibm.com/partnerworld/wps/servlet/ContentHandl
er/POW03017USEN



©
2012
IBM Corporation

Some Words on Memory Virtualization


System Memory is not a dynamic LPAR resource

as CPU cores are.


SAP application memory has several characteristics:


Core memory, linked to a started SAP instance


Executables
, shared segments, buffers etc.




Hard to virtualize


User context


Allocated per users


Fluctuating


Java heap


Managed by Garbage Collector


Fixed allocated per Java node


Consequently, memory is much harder

to virtualize than CPUs




Sizing models of all vendors today allocate m
emory

to virtualized environments, as they do for non
-
virtualized.

22

ISICC
Walldorf
, MK

Start Cldel 514

Start Cldel 504h

Start CLIENTCOPY

TSM Server incr. backup

TSM Server incr.
Backup end

t
0
+110min

t
0

Start SAP

Source: SVA, IBM BP Germany

©
2012
IBM Corporation

23

ISICC Walldorf, MK

AMS recommended usage patterns

Around the World


Partitions support workloads with memory
demands that peak at different times


WW SAP deployments

0
5
10
15
Asia
Americas
Europe
Time

Memory Usage (GB)

Day and Night


Partitions support day time web applications
and night time batch


SAP Dialogue
and Batch instances

Infrequent use


Large number of partitions with sporadic
use


SAP non
-
Production instances


Cluster Standby LPARs

0
5
10
15
Night
Day
Time

Memory Usage (GB)

0
5
10
15
#10
#9
#8
#7
#6
#5
#4
#3
#2
#1
Time

Memory Usage (GB)

©
2012
IBM Corporation

Limited AMS Use Cases


AMS provides limited benefit and is
not

recommended for the following
types of applications:



Workloads with high, sustained memory residency requirements


All the allocated memory is actively referenced


Frequently changing referenced memory regions


Productive SAP systems often have those characteristics








Response time and performance sensitive workloads



Workloads with high degree of load variation



If these workloads are deployed in AMS environment, ensure memory over subscription
is minimal to nil


24

ISICC Walldorf, MK

24hrs

©
2012
IBM Corporation

Customer Feedback on AMS usage in an SAP environment


Performance critical SAP systems need to be protected against memory “underflow”:


Continue to use dedicated memory w/o AMS


With AMS use Loan Policy = Default* with a high memory weighting



Limit AMS memory pooling to less critical, i.e. non
-
PROD, systems


Sandbox
-
, Test
-
, Development
-

and Education (IDES) systems


Set Loan Policy to Default or Aggressive, if AIX paging can be tolerated



20
-

25% overall memory saving compared to aggressive peak
-
sizing


Monitored environment showed aggregated memory distribution as



min = 95GB,
avg

= 113GB, max 138GB, with peak consumption of up to 176GB


Too small memory footprints by sizing can be compensated to a certain degree


Memory peak consumptions are compensated by donated memory from pool


Memory transfer rate into overcommitted LPAR vary:


300MB/sec



from pool was ~20MB/sec in customer environment


Transfer rate from donating LPARs into AMS pool was ~10/MB/sec in parallel



Compared to slow classical DLPAR memory (re
-
) allocation, AMS memory is quickly available to demanding
LPAR


Instead of monitoring and adjusting memory size per LPAR


even in shared LPAR pools


only a single AMS
pool needs to be managed.



Benefit becomes higher the more LPARs dynamically share memory

25

ISICC
Walldorf
, MK

©
2012
IBM Corporation

Active Memory Expansion versus Active Memory Sharing

Active Memory
Expansion


Effectively gives more memory
capacity to the partition



“memory
expansion”


Efficiency depends on compressibility of
in
-
memory content


AIX partitions only


Easy initialization via HMC


Instantaneous effect when
activated



Potential of HW assistance


Permanently requires few CPU
cycles

26

ISICC Walldorf, MK

Active Memory
Sharing


Moves memory from one partition to
another


“memory pooling”


Latency of memory availability


Best fit when one partition is not
busy while another partition is busy


AIX, IBM i, and Linux partitions


Comparably complex to setup


Latency until pooled memory
available to application


N/A


No CPU cycles required after
memory allocation done

0
5
10
15
#10
#9
#8
#7
#6
#5
#4
#3
#2
#1
©
2012
IBM Corporation

Content


POWER7 Processor Characteristics


SMT Modes


Entitlement versus virtual CPUs


PowerVM Advisors



Memory Virtualization


Advanced Memory Expansion (AME
)


Advanced Memory Sharing (AMS
)



PowerVM Live
Partition Mobility



AIX Workload
Partitions (WPARs)



SAP
and PowerVM
Integration


POWER metrics in
CCMS


SAP Cloud Computing



27

ISICC Walldorf, MK

©
2012
IBM Corporation

POWER6 Live Partition Mobility (LPM)

28

ISICC Walldorf, MK

The following DB
-
combinations are
supported by LPM:



DB2



DB2 V9.1 FP4



DB2 V9.5
ff



MaxDB



MaxDB

7.6



MaxDB

7.7



Oracle



Oracle
10, 11.0.2 and
later



Single Instance only, no RAC


AIX 5.3 TL8 SP4, or



AIX 6.1 TL2 SP3


-
AIX 7.1

Movement to

a different server with

no loss of service

Virtualized SAN and Network Infrastructure


As of July ’09 LPM is fully supported by SAP


See
SAP
-
Note 1102760

for latest status, pre
-
requisites, hints and tips

POWER6 Live Partition Mobility Demo with SAP available
:

http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS2921

SAP ACC and Live Partition Mobility Integration Demo available:
http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS4232

SAP license keys

LDAP clients

CPU ID of NIM clients

©
2012
IBM Corporation

LPM Characteristics Quantified

29

ISICC Walldorf, MK

100%

90%

time

SAPS

100%

1 min

2


3 min

Variable Minutes

Estimate =

LPAR size in GB

2…4GB/min


Approx. 5%
throughput
degradation

Hour
-
Glass
appears on
SAPGUI

Approx. 10%
throughput
degradation

Target LPAR

Source LPAR

©
2012
IBM Corporation

Influencing factors for LPM

Values measured during
POWER6 to POWER7
transition at SAP customer site

30

ISICC Walldorf, MK

0 20 40 60 80

100

0

15

30

45

60

75

90

LPAR Size/GB

Elapsed Migration Time/min

LPAR Size

(progressive)

LPAR Utilization

Network

Hops

Whitepaper

Live Partition Migration of SAP Systems Under Load”

available at
TechDocs
:


http://w3.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/WP101917


http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/WP101917



©
2012
IBM Corporation

Practical Experiences with LPM in an SAP Landscape


LPM is being used by many SAP sites meanwhile


This includes non
-
disruptive upgrade scenarios from POWER6 to POWER7



In order to estimate transfer time of a running SAP instance assume an effective transfer rate of




㈠瑯 㐠䝂/浩n u獩ng 愠㸽 ㅇ1 LAN



Users can continue working, no perceived performance impact has been noticed


Users will experience approximately 2…20 seconds of unavailability at SAP frontend during final transfer stage
(hourglass).



SAP License Keys must be available on target system


After LPM transfer SAP Apps continue running, but cannot be restarted


As of SAP NetWeaver 7.3 a Central SAP License Server avoids this problem




Link to
SAP Flexible License Mechanism



Avoid LPM transfer while source LPAR is highly utilized


This will exponentially increase migration time


DBs generate significantly less memory page updates compared to App
-
Servers



Customers have experienced problems with remotely linked SAP systems (via RFCs)


Timing issues occurred during final take
-
over


Test runs required


31

ISICC Walldorf, MK

©
2012
IBM Corporation

Content


POWER7 Processor Characteristics


SMT Modes


Entitlement versus virtual CPUs


PowerVM Advisors



Memory Virtualization


Advanced Memory Expansion (AME
)


Advanced Memory Sharing (AMS
)



PowerVM Live
Partition Mobility



AIX Workload
Partitions (WPARs)



SAP
and PowerVM
Integration


POWER metrics in
CCMS


SAP Cloud Computing



32

ISICC Walldorf, MK

©
2012
IBM Corporation

AIX 6™ Workload Partitions


SAP supports WPARs for production systems.


Shares system WPARs


No resource control


DLPARs (Monitoring)



SAP
-
Note 1105456

describes supported WPAR
environments, prerequisites and restrictions


System WPARs only


IBM DB2 UDB 9.5 + 9.1 FP5 ff


MaxDB 7.6 + 7.7 (incl. liveCache)


Oracle 10.2.04 (no RAC) and later



WPAR Application Mobility not yet supported


No DB supports Appl. Mobility today



Requirements for WPAR monitoring capability in
CCMS


Also documented in above note


CCMS monitoring integration is a prerequisite
for support of production environments.


33

ISICC Walldorf, MK

Workload

Partition

A

Workload

Partition

C

Workload

Partition

B

AIX 6 Image

Workload

Partition

D


Workload

Partition

E


SPLPAR vs. WPAR positioning paper for SAP apps. available at:
http://w3.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/WP101179


single memory space

©
2012
IBM Corporation

Content


POWER7 Processor Characteristics


SMT Modes


Entitlement versus virtual CPUs


PowerVM Advisors



Memory Virtualization


Advanced Memory Expansion (AME
)


Advanced Memory Sharing (AMS
)



PowerVM Live
Partition Mobility



AIX Workload
Partitions (WPARs)



SAP
and PowerVM
Integration


POWER metrics in
CCMS


SAP Cloud Computing



34

ISICC Walldorf, MK

©
2012
IBM Corporation

Challenge

Typical SAP customer landscape:


10..800 SAP Systems


10..800 databases


1..500 servers


3
rd

party software (often on Windows)


Most multi
-
server SAP installations are
heterogeneous

35

ISICC Walldorf, MK

liveCache
liveCache
SAP NW AS
SAP NW AS
SAP NW AS
SAP Appliance
Special
Services
BI
Accelerator
BI
Accelerator
Data
Data
ABAP
Java
ABAP
Java
ABAP
Java
Central Services
ENQ
MSG
Database
DB2
SAP NetWeaver Application Server
SAP Database

Which physical systems are used by my
SAP system?


Are there free resources in my resource
pool? How much can the SAP system
grow?


I need to apply hardware maintenance to
one server. Which SAP systems and
components are affected?

SAP BASIS

IT ADMIN

©
2012
IBM Corporation

Improved PowerVM™ virtualization monitoring within SAP CCMS


In 2006 IBM Power Systems was the first platform to incorporate platform specific (i.e. PowerVM)
virtualization metrics into SAP CCMS. Now additional virtualization capabilities can seamlessly be
monitored using standard SAP transactions (OS07n,...). This is a prerequisite for use of these capabilities
with SAP production instances.



Starting with SAP NetWeaver 7.20 (shipping since 2010) the newly integrated CCMS monitoring
functionalities are
:




Monitoring of POWER shared pools




Monitoring of AIX 6 WPARS (Workload Partitions)


See SAP
-
Note 1105456 "SAP Installations in AIX 6.1 WPARs"



Indication when Live Partition Mobility has been done


See SAP
-
Note 1102760 "POWER6 Live Partition Mobility"



Monitoring of PowerVM Virtual I/O partitions (VIOs)



See SAP
-
Note 1379855 “Installation of
saposcol

and
sapccmsr

on IBM VIOS”



Parameters of Active Memory Expansion and virtual memory size


SAP
-
Note #1464605 titled "POWER7 Active Memory Expansion".




36

ISICC Walldorf, MK

©
2012
IBM Corporation

37

ISICC Walldorf, MK

PowerVM and AIX SAP Monitoring Transaction (ST06, OS07)

Hierarchy of

virtualization layers

now clearly visible

Physical server

LPAR

WPAR

©
2012
IBM Corporation

AME Memory Metrics have been added to CCMS


The memory monitoring capabilities are introduced with
saposcol

version 12.46.


See SAP
-
Note #710975 for more details and updated SAP
-
Note #1464605 titled "POWER7 Active Memory Expansion".

38

ISICC Walldorf, MK

AME specific section

©
2012
IBM Corporation

SAP VIO
-
Server agent for system wide monitoring

39

ISICC Walldorf, MK

POWER Hardware
Hypervisor
Shared Processor Pool
i5/OS
Linux
on
Power
AIX
DLPAR
Virtual Shared Pool 1
AIX
i5/OS
DLPAR
DLPAR
LPAR
LPAR
Virtual Shared Pool 2
Linux
on
Power
LPAR
AIX
DLPAR
Linux
on
Power
LPAR
AIX
LPAR
VIOS
LPAR
Dedicated
Shared
Processor
WLM
System
WPAR
App.
WPAR
System
WPAR
System
WPAR
System
WPAR
System
WPAR
App.
WPAR
WLM
Active Energy Manager
POWER Hardware
Hypervisor
Shared Processor Pool
i5/OS
Linux
on
Power
AIX
DLPAR
Virtual Shared Pool 1
AIX
i5/OS
DLPAR
DLPAR
LPAR
LPAR
Virtual Shared Pool 2
Linux
on
Power
LPAR
AIX
DLPAR
Linux
on
Power
LPAR
AIX
LPAR
VIOS
LPAR
Dedicated
Shared
Processor
WLM
System
WPAR
App.
WPAR
System
WPAR
System
WPAR
System
WPAR
System
WPAR
App.
WPAR
WLM
Active Energy Manager
POWER Hardware
Hypervisor
Shared Processor Pool
i5/OS
Linux
on
Power
AIX
DLPAR
Virtual Shared Pool 1
AIX
i5/OS
DLPAR
DLPAR
LPAR
LPAR
Virtual Shared Pool 2
Linux
on
Power
LPAR
AIX
DLPAR
Linux
on
Power
LPAR
AIX
LPAR
VIOS
LPAR
Dedicated
Shared
Processor
WLM
System
WPAR
App.
WPAR
System
WPAR
System
WPAR
System
WPAR
System
WPAR
App.
WPAR
WLM
Active Energy Manager
CPH

Central Monitoring System

( e.g. Solution Manager

)

Collect OS data from LPARs

with SAP CCMS agents

Reporting TXs


VIOS Certified Program

approval to
run SAPOSCOL and the SAP
SAPCCMSR
-
Agent in the VIO
-
Partitions


The SAP collector and the agent
can be downloaded as an AIX
installp package from the SAP
Service Marketplace


Details about installation and setup
in SAP
-
Note 1379855.

©
2012
IBM Corporation

SAP Launch for IBM Systems Director


Example launch points are SAP
Management Console (ABAP & java
Systems) and SAP NetWeaver
Administrator (Java Systems)



The launch points are automatically
collected from the SAP Solution Manager



The application discovers the SAP Solution
Manager und IBM Systems Director and
creates the SAP Launch points for systems
which are available in both management
applications



The application creates a JSON file on the
file system that should be inserted into IBM
Systems Director with a simple command



The application may be executed
periodically in order to maintain up
-
to
-
data
landscape state


40

ISICC Walldorf, MK




IBM Systems Director

SAP Solution Manager

SAP Launch Points

Application for updating

SAP Launch points

within IBM Systems Director

Collecting data

about available SAP Systems

via CIM protocol

Collecting data

about available hosts

via REST protocol

Creating of

JSON input file

with definitions of

SAP launch points

Upload JSON file

manually via CLI

2

1

3

4

Sample Package with Code and Documentation available for free download at IBM TechDocs:

http://w3.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS4578

http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS4578

http://partners.boulder.ibm.com/src/atsmastr.nsf/WebIndex/PRS4578

©
2012
IBM Corporation

SAP Launch from IBM Systems Director


41

ISICC Walldorf, MK

Uses Advanced External Application Launch (AEAL) integration facility added into Director 6.2

©
2012
IBM Corporation

Integration of POWER Technology into SAP Cloud Computing


The previous section included herein and dealing with PowerVM and SAP
Adaptive Compute Controller (ACC) has been moved to an own
presentation.


SAP Adaptive Compute Controller has evolved to
SAP Landscape
Virtualization Manager (LVM)
in 2011.



See the presentation:

Introduction
to an Integral IBM / SAP Cloud
Solution for SAP
Landscape Management”
at
http://
w3.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS4803


42

ISICC Walldorf, MK

©
2012
IBM Corporation

POWER and SAP

Related Materials



43

ISICC Walldorf, MK

©
2012
IBM Corporation

Helpful „SAP on IBM POWER“ documents

44

ISICC
-

MK, 02/2012


IBM internal competitive + reference materials available as a Cattail Collection:



http://cattail.boulder.ibm.com/cattail/#view=collections/C64753E05C0C3DDA93486644093F23B6


Comfortable Video
-
Demos available at
TechDocs

(includes PW & www):



POWER6 Live Partition Mobility Demo with SAP


http://w3.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS2921



Integration of IBM
PowerVM and SAP Adaptive Computing Controller + CCMS



http://w3.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS4232


Whitepapers available



Live Migration of Power Partitions running SAP Applications (SAP SDN)


https://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/webcontent/uuid/b0b6911f
-
bf10
-
2c10
-
14ba
-
bc789953ff25


SAP Adaptive Computing Controller for IBM Power Systems and IBM System z


http://w3.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/WP101745



Redbook completely reworked in October 2011



SAP Applications on PowerVM


http://www.redbooks.ibm.com/redbooks/pdfs/sg247564.pdf



ISICC Wiki Page covering SAP landscapes on IBM PowerVM


http://w3.tap.ibm.com/w3ki2/display/isicc/PowerVM+and+AIX+Virtualization




©
2012
IBM Corporation

SAP P
rovided Information


Almost all technical SAP materials are maintend on the SAP Service Market Place. It
requires a one time registration (S
-
User) in order to get access to the provided links
and repositories at SAP.



Product Availability Matrix (PAM)


http://service.sap.com/pam


Lists all supported OS/DB/SAP stacks


Another technical Source is SAP Developer Network (SDN):


http://www.sdn.sap.com/irj/sdn


AIX Landing Page in SDN


http://www.sdn.sap.com/irj/sdn/aix




Public SAP Benchmark Entry Page


http://www.sap.com/solutions/benchmark/index.epx


2 tier is most common one and base for SAPS comparisons



45

ISICC
-

MK, 02/2012

©
2012
IBM Corporation

SAP
Notes about IBM POWER and AIX


PowerVM Virtualization


1002461

Support of IBM Dynamic LPAR and Micropartitioning


1102760

PowerVM Live Partition Mobility


1105456

SAP Installations in AIX 6.1 WPARs


1122387

Linux: SAP support in virtualized environments


1464605

POWER7 Active Memory Expansion


Monitoring


994025

Virtualized OS environments in the operating system monitor


113169

CPU utilization metrics of IBM System p


1379855

Include VIOS Partitions into SAP Performance Monitoring


SAP Applications on AIX


1087498

Support for AIX 6.1


1137862

Using SAP systems with AIX 6.1


1096445

AIX 6.1 known problems for validation


1048686

Recommended AIX settings for SAP


1458918

Support for AIX 7.1


1541935

Using SAP systems with AIX 7.1


973227

AIX Virtual Memory Management: Tuning Recommendations


789477

Large extended memory on AIX (64
-
bit) as of Kernel


1088458

AIX: Performance improvement for
ES/SHM_SEGS_VERSION


856848

AIX Extended Memory Disclaiming


1121904

SAP on AIX: Recommendations for paging space


948294

AIX JFS2: mount options to use with Oracle 10g / 11g


1289460

ORA
-
354 on AIX5L if agblksize is not set correctly


46

ISICC
-

MK, 02/2012

For registered S
-
Users

can be accessed via

http://service.sap.com/notes