Trilogy Solution Brief - Retail Performance

8/2/2019 Trilogy Solution Brief - Retail Performance

1/28

RETAIL PERFORMANCE WITH ASSURANCE

we know

they know

RETAIL TRILOgy: A jOINT

IbM ANd SAP PROjECT

FUSINg FUNCTION ANd

INFRASTRUCTURE


2/28

Tale of Contents

3 Introuction

4 backroun to the Retail Trilo Proect

9 Test desin an Scope

15 Results: Hihlihts of Proof-points

21 Choice of IbM Infrastructure for Retail

26 SAP for Retail: A Comprehensive, Interate Solution

27 Conclusions


3/283

Many leading retail companies have recognized the value of implementing an SAP Retail solution on IBM infrastructure. IBM and SAP

have responded to the needs of the retail industry by completing a comprehensive proof of technology for the latest versions of the

mission-critical SAP Retail applications on IBM infrastructure. The results are relevant for small to very large businesses.

High volume point-of-sale data, combined with the increasing requirement for transaction data to be turned into actionable business

information, represent ongoing challenges for retailers of all sizes.

These challenges demand extreme performance and scalability from the supporting applications and IT infrastructure. Additionally,

retailers are looking for reassurance that they can match the application with the infrastructure that best meets their performance,

reliability, and cost requirements.

Together, IBM and SAP initiated a rigorous 12-month multi-million dollar proof of technology project for the industry. Our overall aim

was to demonstrate the fusion of application with infrastructure in a combined solution for retail-specific applications.

Until now, comprehensive and up-to-date performance and configuration information has not been readily available. Without this,

retailers have not been able to optimize their investment for the core time critical elements of the SAP for Retail portfolio spanning

multiple systems on a choice of platforms. These documented tests enable our customers to optimize their investment in SAP for

Retail running on IBM platforms.

We conducted this extensive, rigorous proof of technology, fusing application and infrastructure to:

Prove the scalability of defined SAP Retail industry processes on IBM technology

Optimize the architecture and conduct tests that would provide guidance for detailed sizing and configuration

Optimize the application configuration for common high volume retail scenarios.

This solution brief and the accompanying white paper lay out the background, approach, and results of the study. The results

convincingly demonstrate the viability and value that the fusion of SAP and IBM technology brings to the retail industry. IBM and SAP

are ready to support your organization in assuring your technology investment.

Introduction


4/284

Background to the Retail Trilogy Project

A trilo of SAP components to prove

Every retailer is unique in terms of their data volumes and specific implementation choices around an SAP application footprint, and

yet remarkably similar in many respects. They are confronted by the same requirements to meet deadlines and service level

agreements, and to do so within the constraints of cost-effective IT configurations. Until now, comprehensive and up-to-date

performance and configuration information has not been readily available. IBM and SAP have now provided concrete proof points

using industry relevant business volumes.

SAP provides retailers with industry-leading solutions to assist in all phases of their business. For this set of performance tests, we

focused on the highest volume, most compute-intensive portions of the solution. These are the functional components that produce

the greatest time-critical processing demands, and as a result are where our largest customers most desire extensive testing of the

latest versions of the solution sets from SAP and IBM (Figure 1).

The tests focused on three areas that present time-critical and scalability concerns for our clients, and exhaustively exercised the

related trilogy of components of the SAP solution to discover or confirm configuration best practices that we can now share with ourcustomers:

SAP Point of Sales Data Management application (POS DM) manages the flow of information between store

devices and servers in store, and back office

SAP Merchandising for Retail, which is the Retail version of SAP Enterprise Resource Planning application

(ERP), is the core solution for corporate data and business processes, within the SAP for Retail solution portfolio

SAP Forecasting and Replenishment application (F&R) is the powerful engine that manages the retail supply

chain.

Figure 1: SAP Retail: a comprehensive, integrated solution

Ensure an Inspired

Shopping Experience

Streamline Supply

Chain Operations

Insight Into All Areas Of

Your Business

Anticipate

Shopper Needs

A Solid Foundation for

Corporate Operations

Note: This Solution Brief will use the abbreviations indicated in parentheses above for sake of limited space in graphics.

One additional abbreviation to be used: BW = SAP NetWeaver Business Warehouse (SAP NetWeaver BW)


5/285

dail prerequisites to success

IBM and SAP recognize that retailers are looking for a series of operational assurances in three focus areas at the core of retail IT

processes:

Efficient management of POS data

Effective overnight store replenishment cycles

Timely order management and consistent data.

In order to realize the benefits of your SAP Retail solution, there are a series of daily operations that need to be executed in an

effective and timely manner across dif ferent integrated SAP functional components (Figure 2).

Insiht into Shopper deman requires:

Efficient POS Data Management

Validated consumption data for supply chain planning

Provision of data for analysis.

Anticipate Shopper Nees presupposes:

Control of accurate data and synchronization across systems

Up to date inventory management

Consistent management of prices, promotions and assortments

Timely ordering, including orders from warehouse/distribution centers.

Streamline Suppl Chain Operations with:

Daily forecasting for trends, events and seasonality

Automated replenishment based on stock information, supplier restrictions and factors influencing demand.

Figure 2: Daily prerequisites to success

Anticipate Shopper Needs

Control of accurate data

& synchronization

across systems

Up to date inventory

management

Consistent management

of prices, promotions &

assortments

Timely ordering,

including orders from

warehouse/distribution

centers

ERP

Insight Into Shopper Demand

Efficient POS Data Management

Validated consumption data for

supply chain planning

Provision of data for analysis

POS DM

Streamline Supply Chain Operations

Daily forecasting for trends,

events & seasonality

Automated replenishment based

on stock information, supplier

restrictions & factors influencing

demand

F&R


6/286

goals of the Retail Trilo oint proect

Together IBM and SAP initiated a rigorous 12 month multi-million dollar proof of technology project for the retail industry. Our overall

aim was to demonstrate the fusion of application with infrastructure in a combined solution for retail-specific applications.

The overall goals were to:

Prove the scalability of defined SAP Retail industry processes on IBM technology

Optimize the architecture and conduct tests that would provide guidance for detailed sizing and configuration

Optimize the application configuration for common high volume retail scenarios.

The capacit requirements

In order to measure capacity requirements the test scenarios were monitored and scaled up incrementally to ultimately represent a

retailer with five distribution centers, 2,385 stores, up to 75,000 articles per store and two years of historical sales data. POS Inbound

consumption data represented up to 34.86 million unique article/store combinations, resulting in 17.43 million purchase order lines.

Retailers come in many sizes and forms. The Trilogy project took this variety into consideration. Dif ferent store types and profiles were

created, from small shops with an average of 2000 Stock Keeping Units (SKUs), up to hyper markets with tens of thousands of active

SKUs.

Figure 3 recaps the distribution of number of articles (SKUs) included in each of the 2,385 stores.

Figure 3: Distribution of articles over 2,385 stores

0

20

40

60

80

100

120

0 10000 20000 30000 40000 50000 60000 70000 80000

NumberofStores

Number of Articles per Store

Distribution of Stores via Articles


7/287

Figure 4: Demand-driven supply chain in retail full cycle

Focus on the store replenishment ccle

Given the challenges outlined above, the tests were defined around specific processes. Figure 4 shows the main sequence of steps

involved in managing the store replenishment cycle. Consumption data from stores is aggregated for processing. To exercise the

critical overnight processing, all three major SAP Retail components needed to be tested for scalability and performance:

POS DM

ERP

F&R.

Each of the most demanding steps was tested individually and then the integrated scenario represented the concurrent load across

the SAP landscape during the critical path processing.

Capture

consumer data:

POS Dataprocessing &

Upload

Trigger Store

Replenishment

Planning

Start order

fulfilment:

Create

orders for

internal &

external

supply

Start supply chain

execution: Create

deliveries and

transport orders

Update stocks

across supply chain:

Post goods issues

in Distribution

Centers

Update stocks

across supply

chain: Post

store goods

receipts

POS DM

F&R

ERP


8/288

Critical time winow processin

The substantial workload must be executed within a defined production time window, with system resources optimized to run at peak

efficiency, handling maximum volumes. During this critical phase of processing there is synchronization between POS DM, ERP and

F&R. There is no margin for degraded performance or a system outage that would impact next-day deliveries. Figure 5 illustrates the

processing steps (in orange) that typically take place during the critical time window.

Figure 5: Critical time window overview

06.00 12.00 18.00 24.00 06.00 12.00

Write non-aggregated sales data to

POS DM Tlog tables

Summarise non-aggregated sales data in POS DM

CreateIDo

csinPOSDM

SendIDocsfro

mPOSDMtoERP

SendConsumptiond

atafromPOSDMtoF&R

Process

IDocs

in ERP

ProcessConsumptiondatainF&R

F

RP-RuninF&R

ReleaseandsendOrder

ProposalsfromF&RtoERP

CreateO

rdersinERP

CreateOutboun

dDeliveriesinERP

Post Store

Goods Receipts

in ERP

Post Warehouse

Goods Issues in

ERP

Write non-aggregatedsales data to

POS DM Tlog tables

Summarise non-aggregated sales data

in POS DM

Post WarehouseGoods Issues

in ERP

Time-critical

Not Time-critical

Update

MD, Stock,OP in F&R

Not Time-

critical


9/289

Test Design and Scope

It is possible to conduct performance benchmark tests in laboratory conditions that do not reflect realistic customer scenarios. This

was notthe case for the Trilogy project. Customers can be very confident in the results of this proof of technology as standard

application configurations were used that represent normal, real life, very high volume retail business processes. The testing

confirmed and augmented SAP best practice configuration and system sizing guidelines, which are available to all customers.

Optimize SAP Retail components

Real life scenarios monitore experts

A cornerstone of the joint proof of technology tests was the direct involvement of SAP expert application staff to monitor the

application behavior, and IBM experts to monitor the underlying architecture to identify and correct potential bottlenecks in the

application. This streamlining activity ensured optimum application throughput while maximizing the utilization of the features of the

supporting infrastructure.

It is important to note that the components and configuration throughout our testing represented authorized and supported elements

of the SAP and IBM solutions for retail and were consistent with our joint understanding of best practice for large-scale retail. This was

a test of practical application of the SAP software under common but large scale business conditions not an abstract to prove scale

alone.

Choice of IbM infrastructure

Allowin customers platform choice

The SAP architecture is flexible and runs on all IBM platforms. In practice, retail landscapes often include a mix of differentcomplementary applications and infrastructure that must be accommodated.

Many retail customers are running their SAP solution on UNIX platforms. Often these UNIX platforms are combined with mainframes,

so it was vital to provide valuable insights and guidance to customers of both. This proof of technology spans both the major IBM

infrastructure platforms.

demonstratin how IbM technolo elivers results

The IBM servers and infrastructure deployed were critical to meeting the extreme challenges of these tests. These leading IBM

technologies also provide the most complete capabilities to meet customer non-functional requirements, such as high availability,

continuous operations, accounting compliance, security, automated administration and more.

Showin the inherent scalailit of the ataase

Regardless of platform, IBM database management systems are highly tuned with SAP software for high performance, high

availability and scalability for SAP applications. This tuning is an ongoing activity by IBM and SAP development teams, who provide a

wealth of expertise in matching functionality to application requirements for new industry challenges such as these.


10/2810

Test infrastructure eploe

For Phase One of the tests, all three SAP components were installed on POWER6 595 (see Figure 6). The database DB2 on AIX

shared resources with the application servers. This allowed tests to take full advantage of all the functionality benefits of the solution

stack, notably resource sharing on POWER6, plus data compression and multi-dimensional clustering of the database.

PowerVM enables the flexible allocation of resources to dynamically adjust server capabilities from small to very large environments.

This was the ideal basis for testing the individual component scalability and performance during the Retail Trilogy project and ideal

for load consolidation in production environments. The high-end storage server technology was used with high speed disks and fiber

infrastructure.

For Phase Two of the tests, the overall capacity was increased with the addition of an IBM System z server as database server (Figure

7). The application servers remained on POWER6 595, while the databases were migrated to System z, using the same storage

technology. This allowed the scaling out of the application by freeing resources for additional application server capacity and

consolidated the database servers on System z.

By deploying workload management on System z in its role as an enterprise data hub, the final results built on its strengths to

consolidate distinct workloads and ensure service levels for each application server. The increased scalability of the combined

solution supported the final massive scale out with multiple application servers.

Figure 6: Phase one test infrastructure

Figure 7: Phase two test infrastructure

POS DM ERP F&R

Application Servers on POWER6 595

DB Server POWER6 595

IBM System Storage DS8300 turbo

Phase 1

POS DM ERP F&R

Application Servers on POWER6 595

DB Server System z

IBM System Storage DS8300 turbo

Phase 2


11/2811

Proof point testin

To gather the information necessary to address industry requirements for different size retail customers and verify results for the end-

to-end, integrated SAP solution, a variety of technical tests were defined and executed:

Estalish scalailit component: Good scalability means that processing time can be reduced by increasing

infrastructure capacity, while limited scalability means that processing time is constrained.

Ensure consistenc an linearit component: Ensure that increasing volumes can be processed in a

predictable way over time, otherwise optimize the application. This minimizes production risks and surprises for

customers.

Test hih-loa capailit component : Measure the high volume and massively parallel capability of individual

components. This establishes the maximum throughput given all test resources.

Interate all components on the critical path: Test target volumes for critical path steps in a real business context

by running all process components within the available capacity as in production environments.

A trilo of components to e teste first iniviuallAs the project name reflects, this proof of technology is covering three key components of the replenishment cycle. The project focus

looks at each of the solutions individually and then the interaction of all three which forms the overall business process of

replenishment.

POS dM

As shown in Figure 8, POS DM receives sales data, executes master data checks for stores and products to ensure that only

consistent data is processed downstream, then prepares and aggregates sales information for ERP and F&R.

SAP POS Data Management is specifically designed for high-performance upload of POS data and enables subsequent analysis ofthe data. The key steps selected for the test scope were:

Processing of non-aggregated sales data

Creation of aggregated sales Intermediate Document (IDoc). An IDoc is a container for exchanging data

between SAP components, and between SAP and non-SAP systems

Sending IDocs to ERP

Sending consumption data to F&R.

Figure 8: POS DM architectural overview

BW

ERP

BW

Masterdata

POS

Inbound

Consumption

data

IDocs

Standards

SalesTransactions

VMIXML

Transaction

database

BAPI

Access Module

Aggre-

gate

F&R

POS DM


12/2812

ERP

ERP holds the master records of products and suppliers, prices, locations, and assortments. It also holds all transactional information

required for retail core processes, such as product lifecycle management, assortment control, sourcing and ordering, inventory

management and valuation, order fulfillment and logistics execution. Figure 9 illustrates the process flow for inbound POS data, then

how purchase orders are created and transformed into logistical documents, which will be processed in external warehouses to

ensure products arrive on time and stocks are updated.

The defined test scope covered those tasks which must be processed during the critical path processing window:

Processing of IDocs

Creation of purchase orders from Forecasting and Replenishment order proposals

Creation of outbound deliveries.

Figure 9: ERP architectural overview

Create purchase orders

Create outbound delivery

ExternalWarehouse

ManagementSystem

Post warehouse goods

issue

Stores

Post store goods receipt

Delivery to stores

ExternalVendor

Delivery

to

stores

Standard

order:

Send info

to external

vendor

Send info about

outbound delivery

Stock transfer order:

Create outbound

delivery

Get info about

goods issue

Send info about goods receipt

Process sales IDocs

-> inventory update

-> FI posting

F&R

P

OSDM

Receive sales IDocs

from POS DM

ProcessO

rders

POS

Inbound

Receive order

proposals from F&R

ERP


13/2813

F&R

Supply chain planning solutions need to help retailers to strike the ideal balance between minimizing inventory ownership and

maximizing customer service. SAP Forecasting and Replenishment uses sophisticated algorithms and automation technology to

accurately forecast sales, including promotions and events which influence demand (Figure 10).

Specifically designed for retailers, F&R is made to scale, reach a high degree of automation and reduce user intervention to critical

business situation with its inherent alert system

Scalability and performance testing was performed for all relevant store replenishment tasks which must be processed during the

overnight critical path processing window:

Process consumption data

Run forecasting and replenishment including exception handling, and create order proposals

Transfer of order proposals to ERP

Figure 10: F&R architectural overview

Automatic

Transfer of

Order

Proposals

into Orders

Transfer of

relevant data

Requirement

Quantity Optimization

Order Proposal Release

Management

Exception

Handling

Forecast and

Requirement Calculation

POS DM

PIPE

Manual

Replenishment

VendorVendor Provision of forecast & other data for cooperative scenarios

Master

data

Open

OrdersStock DIFs Sales

F&R

ERP

BW

POS DM


14/2814

The trilo teste as an interate solution

The objective of the integration test was to demonstrate the ability of the complete application and infrastructure solution to handle the

critical time window for very high volumes. This scenario proved the end to end solution, including the interfaces and flow between

solution components. It confirmed the high-volume processing with the concurrent load requirements of each component, consistent

with what is found in a customer production environment.

Figure 11 shows the logical sequence of the integrated process steps and their integration points. The numbering indicates the

sequence of the steps within the replenishment cycle of the integrated retail scenario (including the solution components of POS DM,

ERP and F&R).

POS DM collects and aggregates sales data, and transforms it into IDocs (1) that will be sent to the ERP system (2). There the

necessary stock and value updates will be performed to keep consistent inventory information (3). At the same time the sales data

information will be fed into F&R as consumption time series (4) to form the basis for the forecast calculation and to adjust the stock

information so the replenishment calculation can begin (5).

The processing of sales data IDocs in ERP and the processing of consumption data in F&R run in parallel. Once the replenishment

run is finished, the order proposals are fed back into ERPs relevant buffer tables (6) from where they will be converted into purchase

orders to suppliers and Distribution Centers (7). Once this is completed the logistics execution is kicked off with the creation of the

outbound delivery documents (8). The critical path ends with the processing of the so-called delivery due list.

Steps 9-11, outside of the critical path, complete the full cycle necessary for follow-up inventory postings in the distribution centers

and receiving stores.

Figure 11: Integrated test scenarios

Process consumption data

FRP-Run

Send order proposals to ERP

35M Line Items

Write non-aggregated sales data to inbound queue

Write non-aggregated sales data to TLOG

Aggregate sales data from TLOG

Send consumption to F&R

Create sales iDOCS from aggregated sales data

Send sales iDOCS to ERP

Process IDOCS (update inventory, post FI)

Create Purchase Order & iDOCS

Create outbound deliveries & iDOCS

Post warehouse goods issues

Post store goods receipt

17M Line Items

7M Line Items

17M Line Items

12M Sales Transactions

35M Line Items

35M Line Items

External Vendor

External WMS

POS DM

ERP

F&R

1

2 4

5

6

8

3

7

7M Delivery Items

10M Purchase Order Items

9

10

11 = Sequence of execution


15/2815

Results: Highlights of Proof-points

Summar of results

IBM and SAP have jointly demonstrated the applicability of our solutions for high volume retail. In each area of

the Retail Trilogy project we have significantly exceeded (by factors of 2-3x) previously published or supported

volumes.

We demonstrated performance for the current versions of the SAP application not only the performance of the

individual application components, but through our integrated scenario, the viability of the complete store

replenishment cycle for large scale retail.

Results were not just demonstrated on a single technology, but on a choice of platforms that retailers find

relevant and commonly use.

Utilizing best practice configuration of the SAP application and the features of the IBM database, storage, and

platform we have optimized the joint performance of our combined solution.

Most importantly, we demonstrated this with common, large scale business conditions using only authorized

and supported elements of our combined solution.

Sales ata into insihts

Consistent with the volumes of very large customers, 225 million sales line items were aggregated into ~35 million consumption line

items for processing by other SAP Retail components, other SAP applications, and/or use by other applications, in only 31 minutes.

That consumption data was transferred to F&R, where it was processed in 25 minutes. Thus, both aggregation and processing of the

aggregated ~35 million store location products in F&R was completed in less than an hour. This represented nearly triple the result of

the most previously released information by SAP for POS DM and its integration with ERP.

Harene transaction processin

The processing within ERP, including the data flows between POS DM, F&R and ERP, represented nearly triple the volumes

previously tested.

Forecastin an replenishment to the extreme

Our results demonstrated peak nightly processing of 80 Million item-store combinations (also referred to as product-locations) in less

than 4 hours. This allows large retailers the flexibility to evaluate all store/Distribution Center/item combinations to be executed in a

nightly process if desired, or to continue to segment the processing to support even higher volumes of item/location. This doubled the

volumes previously validated by SAP for the forecasting and replenishment processor run (FRP).


16/2816

Results from the inepenent component tests

During the independent component tests each SAP application workload was tested on a dedicated server with the full volumes of

data in order to establish the maximum throughput possible on a single server. These results are shown in Figure 12.

While providing valuable information about portions of the solution, individual component testing excludes the actual inter-system

dataflow, coordination, concurrent load, and other effects as these cannot be tested in unitary tests. Hence, testing of the complete

business process was accomplished during integrated scenario testing.

Send consumption Data

225M Line Items

31min

Process IDocs

139402 IDocs

3h 26min

Create

Purchase

Orders17.42M

2h 03min

Create

Order

Deliveries6.97M

1h 6min

F&R

POS DM

ERP

Create

Sales IDocs

34.8M LineItems

3min

Process

Consumption

data

34.8M Line

Items

25min

FRP Run

80M product/locations

+Create Order Proposals

17.8M Line Items

3h 56min

Figure 12: Independent Component Tests

Aitional enefit to retailers

As part of the individual component tests we captured information for estimating the system size requirements for the SAP Retail

environment. While we carefully and incrementally scaled the volumes and applied optimization capabilities of the IBM infrastructure,

we monitored and documented the performance of the application environment throughout. This enabled us to construct the most

comprehensive information for the sizing of the SAP Retail application ever available.


17/2817

Figure 13: Integrated scenario processing times

Results from the Interate Solution

We demonstrated in the integrated scenario the viability of the critical path for the complete Trilogy solution for large scale retail taking

into account the inter-system dataflow, co-ordination, concurrent load and other effects inherent in actual retail environments. We

demonstrated the end-to-end process with the same data that we used to demonstrate the scale of the individual Trilogy

components.

Our results in Figure 13 showed that we could process the same high volumes we demonstrated individually, in a stable, scalable,

integrated environment.

Importantl, we emonstrate that this coul e achieve in a winow well within the nihtl processin requirements for most retailers.

In the integrated scenarios, concurrent processing of the critical path delivered the following results:

POS Sales data, aggregated into approximately 35 million consumption line items, processed concurrently by

F&R, and by ERP (for inventory update and financial posting) in under 5 hours.

The final steps of the integrated scenario were the transfer of the optimized order information from F&R to Retail-

centric ERP transactions such as purchase ordering and delivery creations.

The order proposals from F&R received by ERP were converted into 17+ million order line items and nearly 7

million delivery lines in just over 3 hours.

This large scale processing enables retailers to have confidence in the scalability and stability of the combined SAP and IBM solution.

Consumption Data

Prepare and send

225M Line Items

22 minutes

Process IDocs

139402 IDocs

34.8 MLine Items

3 hours 43 minutes

Create

Purchase

Orders

17.42M

2h08 mn

Create

Outbound

Deliveries

6.97M

1h04mn

F&R

POS DM

ERP

Summarizati

on

Create Sales

IDocs

34.8M Line

Items

6 minutes

Process

Consumption

data

34.8M Line

Items

48 minutes

FRP Run

80M product/locations

+Create Order Proposals

17.8M Line Items

3 hours 46 minutes

Transfer IDocs

139402 IDocs

22 minutes

Transfer Order

Proposals

17.8M Line Items

4 hours 22 minutes


18/2818

Optimize infrastructure enales unrivale results

The proof of technology aspects of the Trilogy project have demonstrated that the careful choice of the right infrastructure is critical for

optimal results. In particular, the underlying database and the virtualization technology deployed have a huge impact.

db2 the optimize ataase for SAP

DB2 database features played a crucial role in achieving the performance results. F&R in particular made direct use of the DB2

database features for partitioning and clustering that played a crucial role in improving SAP parallelization of the workload and

enhancing the performance of huge block insert and delete operations. The benefits were significantly increased throughput and

shorter runtimes.

In addition, benefits of features such as automatic storage management, memory self-tuning and software or hardware compression

reduced hardware requirements, which result in an overall better/price performance ratio. This was demonstrated in DB2 LUW by

improved runtimes in our testing with:

56% less database space required, so fewer disks and reduced DB administration time for backup and restore

Up to 66% less Input/Output (I/O) requests, so faster data access

57% less memory utilization, so hardware cost savings

Optimize total cost of ownership throuh virtualization an resource sharin

The more the infrastructure footprint can be reduced while still meeting the application requirements, the less investment is required

and the lower the total cost of ownership (TCO).

The PowerVM shared processor pools allowed multiple SAP systems to be housed on a single server. For the integrated retail

scenario, all three systems share the processor resources according to the requirements of their processing load and according to a

defined priority scheme.

The graph below (Figure 14) depicts the peak requirements of all three systems during an integrated run. This shows the processor

capacity which would be required for dedicated systems. As these peaks can occur at different times, PowerVM allowed us to meet

the workload requirements with only 48% the total number of processors required on non-virtualized systems. This is a tremendous

saving in hardware and software resources. Physical resource sharing allows the optimal use of infrastructure investment.

Figure 14 Peak requirements on dedicated servers

POSF&R

ERP

10

20

30

40

50

60

PhysicalCPUs

Physical Resource Requirements at High Load per SolutionComponent

POS

F&R

ERPPOSF&R

ERP

0

10

20

30

40

50

60

PhysicalCPUs

Peak Requirement per SAP System

POS

F&R

ERP


19/2819

On the database server, System z, we demonstrate the resource sharing for the three concurrent databases. In this design, POS DM

and ERP shared the same virtual machine, as their peak loads occur at different stages in the processing chain. Again we

demonstrate how the peak requirements of nearly 20 Central Processors (CPs) are met by sharing just over 12 physical CPs (Figure

15).

Since the average load on these resources was very low, at any given point in time the remaining processor power was available for

other active work. Work Load Manager (WLM) and Processor Resource / System Manager (PR-SM) ensured the systems priority

scheme matched the companys Service Level Objectives (SLO). As a result we were able to over-allocate processing resources,

confident that we would be able to meet critical response time and batch run time targets.

The combined System z database and POWER6 595 application server architecture demonstrated how the integrated retail scenario

could be implemented to satisfy the strenuous retail industry business SLOs. With this architecture we were able to apply the best

technical characteristics of each system to specific business requirements of the retail solution.

Figure 15: Peak requirements with resource sharing

6

8

10

12

14

#CP

SAP RetailDBs LPARs - CPU utilization - Integrated run

0

2

4

6

8

10

12

14

Max Avg Max Avg

POS DM & ERP F&R

#CP

SAP RetailDBs LPARs - CPU utilization - Integrated run


20/2820

00:00

01:12

02:24

03:36

04:48

06:00

07:12

08:24

09:36

FRP 1 to 2 servers ERP 1 to 2 servers

Highlo

adRuntimesinHours

Scalability - Resources vs Concurrent Load Runtimes 80MilProduct/Location Volume

64CPUs

128CPUs

00:00

01:12

02:24

03:36

04:48

06:00

07:12

08:24

09:36

FRP 1 to 2 servers ERP 1 to 2 servers

Highlo

adRuntimesinHours

Scalability - Resources vs Concurrent Load Runtimes 80MilProduct/Location Volume

64CPUs

128CPUs

Provin scalailit an stailit of the interate solution

Stailit uner test

One proof of stability is to answer the question, what happens as the volume increases? Does the solution remain stable as the

requirements increase?

Here in Figure 16 we proved that the resulting processing times, for the complete integrated business process, remain stable over

time and volume. We started with a medium retail volume and then doubled this volume; achieving this in somewhat less than double

the run-time. The latter also proved the integrated process stability; executing an extended volume over nearly 13 hours.

Scalailit uner test

As the volumes of data grow, the processing window allowed by the business is most likely to remain fixed. We need assurance that

the solution is able to handle increasing volumes of data in the same processing time by scaling out over more resources.

In answer to scalability, we proved that by using additional resources we can reduce the longest running processing steps. During

window of load where both ERP and F&R are concurrently processing the two highest load requirements, we showed that by using

double the application server resources we can cut both the runtimes in half (Figure 17). This demonstrates that the volumes required

can be handled in the window required by the business by applying additional resources.

Figure 17: Double resources to halve the time

1x

Figure 16: Stability over time and volume

1X 2X

Stability: Volume vs Time - Double Volume = Double

Runtime

MedVol

HighVol

00:00

04:48

09:36

14:24

Volumes

IntegratedEndtoEnd

ProcessingWindow

(Hrs)

MedVol

HighVol

1x 2x


21/2821

Choice of IBM Infrastructure for Retail

The Trilogy testing spanned several IBM infrastructure platforms to address the requirements of multiple retail sizes. These included

two POWER6 infrastructures for small and medium business requirements, and two high-end infrastructures, for the large and very

large, which combined the strengths of the IBM System z with POWER6. This huge investment in infrastructure and effort brings

allows us to provide results which are relevant to an industry of many different sizes and requirements.

Scalin up to the IbM mainframe (Sstem z)

Following the complete series of individual tests on POWER6 595, the test included the migration of POS DM, ERP and F&R

databases from POWER6 595 to System z. This is the potential situation some existing customers might find themselves in if they are

not able to meet:

Security demands

Availability requirements

Batch window constraints

Scalability requirements.

All existing POWER6 595 application servers were connected to the System z database server and all resources were retained to

support the scaled-up configuration. Moving distributed databases from UNIX, Linux, or Windows is a safe process for customers,

which is supported by SAP and IBM.

In practice using SAP standard tools built on DB2 for z/OS capabilities it was possible to import, reorganize and compress each

database in one step, then proceed rapidly with the next phase of testing.

Verification of storae eman

By adding a layer of storage virtualization, using the SAN Volume Controller (Figure 18), a flexible growth scenario was demonstrated

and the optimal storage requirements for the different retailer sizes precisely determined. The real life storage requirements are not so

much determined by the amount of data on the disks as the by the behavior requirements of the application.

By scaling out the I/O footprint in relation to the growing demand of the increasing application load we were able to verify the I/O

requirement for throughput and performance. This proof point demonstrates a non-disruptive growth scenario for fiber storage

environments and provides the basis for very precise storage sizing, which reduces risk and aids planning.

Size 1

Size2

Size 3

Virtual

Disks

Size1

Create 10 LUNS

Size 2

Create 20 LUNS

Size3

Create 40 LUNS

LUNs on Physical

Storage Server added

as required

Database

Server

Storage Virtualization

Layer: SAN Volume

Controller

Increasing IO

requirement

As DB scales up

Storage Virtualization with

Non-disruptive scale-out

to meet increasing load

Figure 18: Storage virtualization


22/2822

ERP AS

POS DM AS

F&R AS

ERP CI+DB

F&R CI+DB

POS DMCI+DB

POWER6 595

64 CPUs 5GHz

SAN

DB2

V9.5

F&R

DB2

V9.5

POS-DM

DB2

V9.5

ERP

DS8300 Turbo

Dual Frame

80TB

virtualEthernetBackbone

Infrastructure that est meets requirements

Incuation test confiuration

The incubator is an IBM hardware landscape installed within the SAP development computer center in Walldorf, Germany, for ease

of access by SAP development. The architecture shown in Figure 19 was based on a midrange POWER6 system and a midrange

storage server to provide a reference architecture for small- to medium-size retail customers.

This is where the systems were first implemented and functionally tested. Each solution was similar to a customer implementation

insofar as the data design and layout was intended to replicate real-life business processes. Data loading was to prepare for high-

volume tests and historical data was generated for two years of sales data on a weekly level, 30 days on a daily level.

Phase One Mass test confiuration

Following validation of the test scenarios on the incubation system in Walldorf, the complete test environment was transferred to the

IBM Customer Solution Center in Montpellier, France. The Phase One Mass test architecture is shown in Figure 20.

This high-end POWER6 landscape consisted of the newest model of the IBM UNIX servers, POWER6 595 64-core 5GHz with 512GB

of memory in collaboration with a high-end storage server DS8300 with 128GB cache and 384 available disks, configured in RAID 5.

In this configuration, where the application servers and DB servers are all housed on a single 595, POWER6 processor sharing is

used to allow the system resources to follow the processing load across the integrated landscape. POWER6 supports multiple shared

processor pools which were used to implement a firm priority system and controlled distribution of resources, so that the databases

and SAP central instances had priority.

Figure 19: Incubation Figure 20: Phase One Mass test configuration

POS DM

ERP

virtualEthernetBackbone

F&R

POWER6 570

16 CPU 3.5GHz

SAN

DB2

V9.5

F&R

DS5100 18TB

DB2

V9.5

POS DM

DB2

V9.5

ERP


23/2823

Interate test confiuration

For the high-end critical path tests, the architecture was extended to the heterogeneous configuration shown in Figure 21. The

databases were migrated to DB2 for z/OS using workload management to enforce a further priority scheme, while the resource

sharing implemented for the pure POWER6 landscape was carried forward into the final tests.

Maximum test confiuration

This final test represented the very high-end volume for an extremely large retailer with a tight processing window for store

replenishment activities, so for the high-load phase of the parallel processing an additional 595 application server was added as

shown in Figure 22.

Figure 21: Integrated test configuration

DS8300 Turbo 40TB

ERP AS

POS DM AS

F&R AS

ERP CI

F&R CI

POS DM CI

POWER6 595 64 CPUs

SANDB2

POS

DB2

F&R

DB2 ERP

POS DM DB

ERP CI

F&R AS

z10 E56 50 CPUs

Figure 22: Maximum test configuration

DS8300 Turbo 40TB

ERP AS

POS DM AS

F&R AS

ERP CI

F&R CI

POS DM CI

Two POWER6 595 64 CPUs

SAN

DB2

POS

DB2

F&R

DB2 ERP

POS DM DB

ERP CI

F&R AS

z10 E56 50 CPUs


24/2824

deliverin unique capailities with IbM technoloies

This project is another example of the shared goal of enabling efficient integration between SAP applications and IBM servers. The

proof points demonstrated how each IBM component in the test infrastructure played a crucial part to achieve the required results.

POWER6 Sstem

IBM Power System servers are the leading choice for enterprise SAP customers on UNIX because of their built-in performance and

virtualization features, as well as their reliability, availability and serviceability (RAS) characteristics. This platform is the leading RAS

solution within the distributed server community.

Processing demands continue to grow, so performance matters. IBM Power systems deliver leading benchmark results for SAP

solutions with fewer processors than competitive systems. The ability to concentrate more SAP instances onto fewer servers while

absorbing increased workloads without corresponding increases in server capacity, offers the potential to achieve new levels of cost

effectiveness. Virtualization is the key technology trend in SAP system infrastructures and this project demonstrates just how IBM

Power systems meet scalability demands while delivering exceptional performance.

IBM PowerHA for AIX (formerly High Availability Cluster Multiprocessing HACMP) provides a mature clustering solution,

commonly used in SAP implementations to protect the SAP databases and application servers from outages; planned or unplanned.

Sstem z

System z delivers extreme business value through industry-leading security, availability, scalability, virtualization, and management

capabilities, components of which have been inherited by POWER6 595. System z can simplify and reduce costs through

consolidation with a high degree of automation, plus offering leading high availability solutions to limit the business impact of SAP

downtime such as reduced profits and lost productivity.

System z is able to consistently maintain high levels of system utilization (90-100%) while ensuring that high priority businessprocesses meet service level objectives (SLO). This characteristic reduces power consumption, and footprint requirements, and

allows an enterprise consolidation which reduces operational requirements thereby reducing total cost of ownership (TCO).

By running the SAP databases on System z for the integrated retail scenario, we were able to demonstrate the vital role of System z for

very large scalable SAP databases based on these features:

Hardware data compression

System offload with the z Integrated Information Processor (zIIP) for database processing

Extended accounting capabilities

Workload Management (WLM) enabled a higher priority for database processing of defined processes.

This proof point also demonstrated how each IBM component in the test infrastructure played a crucial part in achieving the required

results.

IbM storae

IBM System Storage DS8000 solutions are designed to deliver robust, flexible, highly available, reliable and cost-effective disk

storage to support continuous operations for large enterprise and mission-critical workloads. They offer improved data protection

through striped disks with dual parity (RAID 6) which can recover from the loss of two disks.

IBM products help organize and manage a storage environment that can better accommodate explosive data growth while

simplifying administration and addressing rising costs. IBM offers traditional hard disk storage as well as solid state disk storage forhot spot data structures.


25/2825

SAN Volume Controller (SVC)

SVC is used to virtualize the storage, which increases utilization, handles peak loads and avoids performance bottlenecks very

efficiently. It also reduces total cost of ownership (TCO) with reduced power and cooling expenses while streamlining administration.

In this project, SVC was used for the most throughput demanding portion of the trilogy F&R, as shown in Figure 23. It was possible to

easily and quickly move data online to different groups of disks (few, many, lots of). We could then prove that a defined number of

disks suffice to handle a given load and also to create a precise storage sizing profile. Besides deploying SVC to better manage the

normal situation of growing data volumes, customers have increased flexibility, for example in order to migrate the production

environment to newer storage technology.

db2

IBM database systems on each platform are optimized for SAP software and highlights include:

Simplified and tailored setup

Performance gains at lower cost

Data compression (software or hardware)

Index compression

DB partitioning

Online backup

One stop service and support via SAP processes.

Parallel database with DB2 data sharing on System z is a unique SAP certified scale-out solution that allows zero downtime

maintenance and upgrades of the DB management system itself.

SAN Volume Controller

Managed Disk

Group 1 (80 Disks)

Managed Disk

Group 2 (160 Disks)

Managed Disk

Group 3 (320 Disks)

Virtual

Disk

Medium

load

Max

Load

Small

Customer

F&R AS

F&R CI + DB

Few Disks

Many Disks

Lots of Disks

VirtualEthernetBackbone

DS8300 Turbo

Dual Frame

POWER6 595

64 CPUs 5GHz

Figure 23: F&R mass test with SVC


26/2826

SAP for Retail: A Comprehensive, Integrated Solution

The SAP for Retail solution portfolio enables you to get ahead, and stay ahead of your competition. By gaining deeper understanding

of your shoppers and your operations, and leveraging superior operational control, you will maximize the effectiveness of your

merchandising strategies.

Insiht into all areas of our usiness: Enable an extensive understanding of all aspects of your operations from

shopper demand through all financial, merchandising, supply chain and store performance that can be

accessed and leveraged across your enterprise.

Anticipate our shoppers nees: Achieve profitable growth through timely execution of strategic merchandise

decisions, including merchandise and assortment planning, lifecycle pricing (through inventory clearance) and

promotion management.

Streamline suppl chain operations: Ensure the right inventory is in the right location, and react faster to changes

in the market through agility, transparency, and global capabilities throughout the supply chain.

Ensure an inspire shoppin experience: Build a compelling and consistent shopping experience across all

channels you serve to entice your shoppers to return again and again. Use best business practices to achieve

unparalleled operational efficiency while maintaining security and minimizing risk.

A soli founation for corporate operations: Streamline financial and human resource activities to maintain

compliance, deliver shareholder value, exceed employee expectations and become a best-run business.

SAP for Retail contains solutions that address the needs of all stakeholders and line-of-business owners across the entire retailing

spectrum. Following is a representative, partial, list of solutions:

SAP Merchandising for Retail

SAP Merchandise and Assortment Planning for Retail

SAP Loyalty Program

SAP POS for Retail and SAP Enterprise POS for Retail

SAP POS Loss Prevention for Retail

SAP Price Optimization for Retail

SAP Promotion Management for Retail

SAP Markdown Optimization for Retail

SAP Multichannel Order Management for Retail

SAP Web Channel Enablement for Retail

SAP Extended Manufacturing for Retail

SAP Extended Warehouse Management for Retail

SAP Extended Procurement for Retail

SAP Centralized Electronic Funds Transfer

SAP Transportation Operations for Retail

SAP Workforce Management

SAP Learning Solution

SAP Real Estate Management


27/2827

Conclusions

The IBM infrastructure fusion, with the SAP Retail application provides a solution for retailers of all sizes. This rigorous deep dive into

the Trilogy components and the requirements of the integrated solution in a near real-life implementation prove the most demanding

retail volumes can be achieved.

Strateic clarit

With its comprehensive set of integrated business solutions, the SAP for Retail portfolio gives you the means and tools to drive your

strategy through to execution. It will help you understand your shoppers preferences and react to demand while providing real-time

visibility across your organization to pro-actively manage your business. This project demonstrated that the SAP for Retail portfolio

operating on a choice of IBM infrastructure supports the critical retail processes involving extreme data volumes that need to be

processed during demanding overnight time constraints.

Increase assurance

The joint Retail Trilogy project demonstrates that retail-specific business processes are well understood. The performance tests and

system landscape deployed accurately reflect retailer needs, while the high volumes and configuration options represent real

requirements. As a result, sizing recommendations are validated for medium and large customers while allowing for growth; and

customers can have confidence in well-justified total cost of ownership (TCO) scenarios for various IBM and SAP solution options.

Maximum performance at optimum cost

This project demonstrated that the time-critical overnight processing can be completed within defined time windows for both normal

and extreme volumes on a cost-effective configuration. IBM technologies including virtualization, DB2 database and enterprise

storage systems were essential to meeting these demands. The test results enable more precision in planning retailer configurations

and take into account the optimal resource utilization achieved on IBM platforms. In addition, IBM platforms are designed to deliver

the highest levels of reliability and serviceability, while offering solutions for the most demanding availability requirements.

Operational excellence

With their combined industry expertise IBM and SAP can ensure the business efficiency of your retail solution. SAP for Retail solutionsprovide the essential business functions across the entire value chain from manufacturers through suppliers and retailers to

consumers, with optimized processes and controls. IBM has been part of every significant technology innovation in retail, including

point-of-sale systems, bar code and radio-frequency technology, as well as e-commerce.

Reassurance from SAP an IbM

The results convincingly demonstrate the viability and value that SAP and IBM technology brings to the retail industry. IBM and SAP

are ready to support your organization in assuring your implementation.


28/28

Trilogy Solution Brief - Retail Performance

Documents