ibm's recommended approach for optimizing …s recommended approach for optimizing different...

IBM's Recommended Approach for Optimizing Different Kinds of SAP Workloads

Agenda

Different workloads

Options for implementations

Moving workload between the layers

2

Workloads and Similarities

What is the universal vehicle ? Vehicles

ground waterair

Cargo planePrivate jetHelicopter

Glider...

TruckFormula-1Family carSports car

...

Cargo shipCruiser

BoatCanoe

...

���� There is no individual vehicle available to fulfill all requirements everywhere and anytime

3

4

Workloads in the IT World

Addressing challenges with unique capabilities on H W and SW level

Transaction Processing

Operational Analytics

Mobile Data Serving

Reporting and Analytics

5

Why IBM ?

• IBM is industry-leading and uniquely positioned to design/execute/implement a Big Data strategy:

– storage (DS8K, V7000, Flash, …)– server (POWER, System z, … )– appliance (Pure Flex, Pure Application, …)– software (DB2, Cognos, Tivoli, Websphere, …)

6

Next Generation Database = ONE 4 ALL

Transactional Analytical Near-line Storage

SAP Business-Suite,Industry Solutions

OLTP workload

SAP BW Big Data

OLAP workload

NLS for BW

Near-line Storage

DB2 LUW DB2 LUW DB2 LUW

Agenda

Different workloads

Options for implementations

Moving workload between the layers

7

Source database: 8 TB on BW 7.0 (non-Unicode)

2-tier implementation: database and application server are sharing the resources

Low hardware required for large SAP BW systems

8

IBM HW x440- 40 cores & 768 GB RAM- without storage (available from customer)

Estimated HW Infrastructure for Production

Assumption yearly 20% growth

8

9

New technology for accelerating queries• CPU-optimized unique runtime handling

• Unique encoding for speed and compression

• Unique memory management

• Vector processing (e.g. SIMD)

• Built directly into the DB2 kernel

DB2 capacity in Peta Bytes and more• One partition: 2,048 PB (uncompressed approx. 20,000 PB)

• One partition: 128 x 1010 rows per table

DB2 10.5: Acceleration for SAP Applications and Big Data Era

SAP BW scenario• Fact table with 438 Million records (~200GB)

• Size (table + indexes)

- DB2 10.1 compressed: 33 GB

- DB2 10.5 compressed: 12 GB

• DB2 parallel query degree switched ON

DB2 10.5 Test environment• 2 Intel® Xeon® processors (8 cores in total), 2.14GHz

• 32 GB RAM, SLES Linux V11

• SAP BW 7.30

10

Test 1: 15 GB Bufferpool

Test 2: 7.5 GB Bufferpool

1744.17

7.09

2790.3

7.35

0

1000

2000

3000

Test 1 Test 2

DB2 V10

DB2 BLU

246

380

DB2 10.5: Performance Improvement for SAP BW

10

Agenda

Different workloads

Options for implementations

Moving workload between the layers

11

Workload distribution and Impacts of Program Code

• Run time consists of- front-end and back-end time- database time- application time

• Application programming and QUALITY is very important for the overall run time

• Example: Carl Friedrich Gauss (1777–1855) Sum of numbers up to 1001 + 2 + 3 + 4 + ...... + 100 = ?

100%

12

Application Layer (~80% of cores and RAM)

Database Layer (~20% of cores and RAM)

• keep workload, where it fits best

• tailored building blocks for highest flexibility

• tailored performance on OS / DB / storage level

Application data is in RAM

2-tierarchitecture protects the performance between Database + application

... DB2 supports 2-tierDatabase data >98% in RAM

Workload shift from application to database layer ?

13

within server

DB

and

App

l Ser

ver

with

in „

neig

hbor

LP

AR

“

SAP Client Copy Runtimes (2-tier vs. 3-tier)

14

• 2-tier < 3-tier comparison with different configuration patterns

• Para 3-tier is almost the best compromise for performance vs. operation and license cost optimization aspects

15

SAP SD Benchmark Comparison

15

IBM‘s Recommended Approach

1. Use the best matching server/storage configuratio n• Appliance solutions, if the workload is the same an d not volatile

• Tailored solutions, if the workload is different an d volatile

Hint: Avoid to find the matching problem for an available solution. The solution is there to solve the problems

2. Use the matching tier-architecture• 2-tier architecture for performance sensitive envir onment

• 3-tier architecture for de-central implementation a nd separation of the layers

Hint: Avoid to include a de-central implementation approach into a 2-tier layer by force. If helpful than install a 2.5-tier layer

3. Use the matching SW technology • OLTP implementation for large number of users with short running transactions (e.g. SAP Business Suite )

• OLAP implementation for large amount of data spread over fewer users (e.g. SAP BW)

Hint: Avoid to melt OLTP and OLAP workload, because of the limitation of appliance solutions

16

• Join me in the Expert Networking Lounge at Date/Time •(if appropriate, depending on your presentation time)

• Stop by the IBM booth . -IBM experts will be on hand to share IBM's extensive portfolio of breakthrough technologies, innovative solutions and comprehensive services for SAP clients.

- Ask us about cloud computing, enterprise mobility, IBM solutions for SAP applications, and social tools for business and commercial engagement.

- See demos and user cases from the IBM’s Client Center: Lab for SAP Solutions.

• Find customer case studies and more on the IBM-SAP Alliance website : www.ibm-sap.com

Further information

17