DB2 11 and 12 for z/OS Performance Update Baltimore Washington DB2 RUG

Akiko Hoshikawa 12/14/2016

Please note IBM’s statements regarding its plans, directions, and intent are subject to change

or withdrawal without notice and at IBM’s sole discretion.

Information regarding potential future products is intended to outline our general

product direction and it should not be relied on in making a purchasing decision.

The information mentioned regarding potential future products is not a commitment,

promise, or legal obligation to deliver any material, code or functionality. Information

about potential future products may not be incorporated into any contract.

The development, release, and timing of any future features or functionality described

for our products remains at our sole discretion.

Performance is based on measurements and projections using standard IBM

benchmarks in a controlled environment. The actual throughput or performance that

any user will experience will vary depending upon many factors, including

considerations such as the amount of multiprogramming in the user’s job stream,

the I/O configuration, the storage configuration, and the workload processed.

Therefore, no assurance can be given that an individual user will achieve results

similar to those stated here.

12/13/2016World of Watson 2016 2

Performance Updates

DB2 H/W Synergy Updates Recent Performance PMR Experience

Performance Topics in DB2 11

DB2 11 Performance Features in DB2 11 New Function Mode

Performance Topics in DB2 12System level Performance Feature

3

Agenda

Challenges

• Ever increasing transactions and data. Can DB2

handle?

• Yet, we need to reduce IT costs

4

”Out of the Box” CPU saving

added first 64 bit support

What Have We Done?

No more low

hanging

fruits

Finished 64 bit

conversion

PCI – Processor Capacity

Index (IBM MIPS)

141-way

Customer Processors

PCI for 1-way

1695

Memory

10 TB

System I/O Bandwidth

832 GB/Sec*

80-way

64-way

54-way

1.5

TB

512

GB1202902600

288 GB/sec*

172.8 GB/sec*

15143 TB

384 GB/Sec*

101-way

z10 EC

z9 EC

z196

zEC12

z13

IBM z13

Will larger buffer pool really help my workload?

• Depend on,

• Access pattern

• Active working set size

Buffer Pool Simulation - DB2 11 CM with PI22091

• Benefit from increasing size of buffer pools varies depending on the workload

• Data access pattern, current buffer pool configuration, etc.

• Provides accurate simulation results from increasing buffer pool size as you run your production workloads

control blocks for simulated pools

Simulation pools = SPSIZE

Buffer poolcontrol blocks

Buffer pools = VPSIZE

How To Use Simulation – Example

• Simulate BP1 additional 500,000 for BP1 with seq. prefetch threshold of 30 % instead of current 60 %

• Collect statistics data or DISPLAY BUFFER POOL DETAIL command outputs • Recommend to reset SPSIZE(0) when simulating different SPSIZE

• Recommend to take enough samples (2-3 hours ) per SPSIZE

• Cost Storage : SPSIZE * 0.02, CPU: approx 1% per Simulated Pool

- ALTER BPOOL(BP1)

SPSIZE(500000) SPSEQT(30)

- DISPLAY BUFFER POOL (BP1) DETAIL

continue to run your workloads

(>1hour)

- DISPLAY BUFFER POOL (BP1) DETAIL)

Measurement IntervalUse this output as simulation result

Simulation Output

DSNB432I -CEA1 SIMULATED BUFFER POOL ACTIVITY -

AVOIDABLE READ I/O -

SYNC READ I/O (R) =25463982

SYNC READ I/O (S) =81181

ASYNC READ I/O =15470503

SYNC GBP READS (R) =11172099

SYNC GBP READS (S) =4601

ASYNC GBP READS =1181076

PAGES MOVED INTO SIMULATED BUFFER POOL =53668641

TOTAL AVOIDABLE SYNC I/O DELAY =35321543 MILLISECONDS

Display output from –DIS BPOOL (BP1) DETAIL or statistics trace class 1, 2 for the measurement interval

Sync I/Os / interval = avoidable sync I/O per second • (25463982 + 81181)/360 = 70958 I/O per sec

Customer Example : Avoidable I/O by Member

0

1000

2000

3000

4000

5000

6000

7000

DP30 DP31 DP32 DP33 DP34 DP35 DP36 DP37 DP38 DP39 DP3A DP3B

Avoidable Sync I/O per sec

1X

2X

5X

10X

20X

40X

System z Synergy Features

12

zHyperWrite Log Write Response Improvement with PPRC

zHyperWrite function for DB2, z/OS and DS8870 with GDPS or TPC-R HyperSwap

• Response time reduced up to 58%

• Benefit percentage varies with distance

Requirements • z/OS and DFSMS APARs

• DB2 10 and DB2 11 SPE APAR PI25747

• Set ZPARM (REMOTE_COPY_SW_ACCEL)

• IBM DS8870 Storage Subsystem MCL

• R7.4 87.4x.x.x - DS8870 242x model 961

DB2 Response Reduction with zHyperWrite

050

100150200250300350400450

Elapsed Commit Wait

tim

e in

se

c

HyperWrite Performance in Data Sharing / Multi Threads

Base PPRC HyperWrite

-22%

-40%

DB2 concurrent insert batch jobs (commit every 10 insert) in data sharing shows 40% reduction in commit response time and 22% elapsed time reduction

Additional CPU in master SRB for triggering additional I/O request

DB2 11 zIIP eligible

z13 Simultaneous Multithreading (SMT)

• SMT-2 allows instructions from two threads to execute on zIIP or IFL core

• Designed to deliver overall capacity (throughput) improvement

• Capacity gain per core will vary • z/OS team’s benchmarks observe 10 to 40% capacity

increase

• With SMT, each threads run slower but more tasks can be dispatched

• Available for tasks running under zIIPs• System parm MT_ZIIP_MODE=1|2 – active threads per online

zIIP core

15

80 50

SMT-2 Eligible = DB2 zIIP Eligible

• DB2 DRDA

• DB2 Query parallelism (parallel tasks)

• DB2 Utilities (selective phases and sort)

• DB2 System tasks running under enclaved SRBs• DB2 10 (prefetch reads and deferred writes)

• DB2 11 (most of async SRB works including in DBM1 and MSTR address spaces)

• DB2 XML parsing

• JCC T2 Application (JVM)

16

• SMT-2 shows throughput improvement of up to 20% with DRDA workload compared to SMT-1 on z13

• No elapsed time impact observed in DRDA/OLTP workload

• zIIP capacity increase

• General CPU time reduction when there are not enough zIIPcapacity

• IIPCP > 0

• Benefits likely depend on the type of workloads and configuration

• With C4Z4, zIIP Capacity factor was 1.2

17

DB2 DRDA Workload (IBM Brokerage OLTP)

(Controlled measurement environment, results may vary)

Transaction CPU time (ms)

Throughput (ITR – trx/sec)

Shared Memory Communications-Direct Memory Access (SMC-D)over Internal Shared Memory (ISM)

• zEC12 introduced Shared Memory Communications over RDMA (SMC-R)

• Using RoCE express feature

• z/OS LPAR to z/OS LPAR, could be across CECs

• Z13 GA2 introduces Shared Memory Communications-Direct Memory Access (SMC-D)

• Direct LPAR to LPAR communications within CEC

• Very similar to SMC-R (over RoCE) extending the benefits of SMC-R to same CPC operating system instances without requiring physical resources (RoCEadapters, PCI bandwidth, NIC ports, I/O slots, network resources, 10GbE switches etc.).

18

Preliminary measurements with SMC-D(z/OS to z/OS)

19

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0

0.001

0.002

0.003

0.004

0.005

0.006

0.007

0.008

SMC-D HiperSockets 10Gbit ethernet

thro

ugh

pu

t p

er s

ec

tim

e in

se

con

d

DB2 CL.1 elapsed time DB2 CL.1 CPU time ETR (trans per second)

Recent Performance PMRs (1)

• Real Storage Management = AUTO/OFF/ON update by PM99575• Possible CPU increase with RSM=AUTO in DB2 master address space or DRDA

threads due to overhead of releasing real storage in z/OS

• If there are enough memory, or repeating same set of packages per thread, RSM=OFF will provide performance benefit

• If you do not have enough memory, RSM=OFF could possibly cause memory increase

• Log read through replication products • Possible CPU increase due to CLAIM request increase in DB2 11

• Solution : PI70095 design change not to use CLAIM

• Reducing Buffer pool size • Takes large amount of DBM1 CPU time to contract VPSIZE

• Working as designed, but we will look into20

Recent Performance PMRs (2)

• Slow to insert into PBG (APPEND YES) from 800 concurrent clients under non data sharing • Due to space map contention at the end of TS

• Using MEMBER CLUSTER and increasing the page size to 32K solved the problem

• DB2 12 addresses this

21

Performance Updates

DB2 H/W Synergy Updates Recent Performance PMR Experience

Performance Topics in DB2 11

DB2 11 Performance Features in DB2 11 New Function Mode

Performance Topics in DB2 12System level Performance Feature

22

Agenda

23

Extended LRSN/RBA and LRSN Spin Avoidance – NFM

• LRSN spin can be totally eliminated in DB2 11 NFM

• Requires both BSDS and objects to be extended log format to see the benefit• Log record size increase starts the point of BSDS conversion • DB2 12 requires BSDS conversion but not for catalog/user objects

-33%

-47%-15% LRSN spin is

recorded in DB2 statistics service fields

QJSTSPNN

QJSTSPNI

Expect approx.

50bytes log record

increase at BSDS

conversion

24

Not Logged DGTTs - NFM

NOT LOGGED option (default LOGGED) • ON ROLLBACK DELETE ROWS (Default, rows deleted) • ON ROLLBACK PRESERVE ROWS (rows preserved)

• Less CPU from log processing, less log volume • Faster rollback/error processing after large insert into DGTTs • Application programmers must take ROLLBACK behavior into consideration

0 10 20 30 40 50 60 70 80 90

Elapsed

CPU

Elapsed

CPU

Elapsed

CPU

Elapsed

CPU

Inse

rtU

pdat

eD

ele

teR

OLL

BA

CK

Time in Second

Without LOG With LOG

2.8x Elapsed Reduction on ROLLBACK

Indirect Reference (Overflow Records) - DB2 11 NFM

Indirect References (overflow records)

– Created during UPDATE against variable length rows or compressed rows

row-1row-2 row-3

row-100row-101

row-2Pointer to row-2

Impact caused by indirect references – Additional getpages, potentially additional I/Os to the overflow pages– Lower clustering– REORG TS is necessary to remove indirect references

Update row-2with larger data

Data Page 1 Data Page n

Reduction of Indirect References - NFM

• Insert process to reserve the space for subsequent update

CREATE TABLESPACE TS1

PCTFREE 20 FOR UPDATE 10

• DDL CREATE/ALTER with PCTFREE x FOR UPDATE y • x = % of free space to leave in each data page by LOAD or REORG • y = % of free space to leave in each data page by INSERT, LOAD or REORG• INSERT will preserve y% while REORG will preserve (x+y) %

• System parameter PCTFREE_UPD (AUTO, 0 to 99) • System default 0 : same behavior as DB2 10

• PCTFREE_UPD = AUTO and FOR UPDATE -1• DB2 determines the value using the history of UPDATE behavior based on Real Time Statistics

(UPDATESIZE is added in RTS) • Recommendation : Use FOR UPDATE -1 unless you know better

• Applicable for newly created objects as well as existing objects (PI12400)

Example : IBM Relational Warehouse Workload

Performance Updates

DB2 H/W Synergy Updates Recent Performance PMR Experience

Performance Topics in DB2 11

DB2 11 Performance Features in DB2 11 New Function Mode

Performance Topics in DB2 12System level Performance Feature

28

Agenda

DB2 12 Key Features Sy

stem

Per

form

ance •Exploit in

memory optimization

•Contiguous BPOOL

• Insert Algorithm

•Better Commit LRSN

•More zIIPsupport

SQL

Perf

orm

ance •Dynamic Plan

Stability

•Runstatsprofile

•Query CPU improvement in Union All, Outer Join

•Adaptive Index

•UDF Caching

An

alyt

ics,

Clo

ud

an

d M

ob

ile A

pp • IDAA

•RestAPI

•Paginations

•Piece meal wise delete

•XM, JSON improvement

•Complex Triggers

DB

A a

nd

24

*7

•Easy Migration

•Online Schema

•PBR RPN

•LOB zEDCCompression

•Security

•Async Lock Duplexing

•Profile improvement

DB2 12 Key Features – System Performance Sy

stem

Per

form

ance

• Exploit Large Real Memory • Index in memory =Fast Traverse Block • Contiguous buffer pool, PGESTEAL(NONE)

• zIIP Usage • Scalability improvement

• Insert Algorithm 2 • Latch contention reduction• EDM pool management improvement

• Better lock avoidance with long running UR

•

Index to Data: Random Keyed Access

Index IX1 with 5 levels

Fast Traverse Block contains non-leaf pages

Limitation : Unique index with size of 64 bytes or less

Turned on as default with upper limit control by user Index Memory Optimization Daemon process monitors index usage

Object level control possible through catalog table

Table

N4

N3

N2R

In Memory Structure (Fast Traverse Block)

How FTB Works ?

• Turn on/off and upper limit controlled via DB2 system parameter INDEX_MEMORY_CONTROL • AUTO, DISABLE or 500-200000 MB

• Using the default, DB2 uses up to 20% of sum of allocated buffer pools

• DB2 monitors candidate index usage and select the indexes to create the in-memory structure

• Main metric is traverse counter with other adjustments

• Users can influence the behavior through catalog table input SYSIBM.SYSINDEXCONTROL, similar to pseudo delete entry cleanup

• New display command to monitor the objects and storage usage

• New instrumentations added to record FTB usage

Using Index in Memory – Batch SELECT

6%

11%

16%

23%

2 3 4 5Index Levels

Improvement (%) from Simple Lookup with DB2 12 vs. DB2 11

0 1 2 3 4 5 6 7 8 9 10

OLTP-simple5 (IMS/DB2)

OLTP-simple2 avg (Distributed)

OLTP-simple1 (IMS/DB2)

OLTP- simple3 (CICS/DB2)

OLTP- complex (SQLPL)

OLTP- simple4 (SPROC)

OLTP CPU(%) DB2 11 vs 12

FTB=AUTO FTB=Disable

zIIP Updates

• CPU parallel queries - up to 80% -> up to 100% zIIP eligible

• REORG and LOAD utility RELOAD phase

36

141

52

60

147

V11 V12

CP

U IN

SEC

ON

D

REORG UTILITY CPU USAGE FROM ESP CUSTOMER’S DATA

zIIP

CP

Introducing New Insert Algorithm

37

Page A

Thread-2

Thread-3

Thread-1

Thread-1

Thread-2

Thread-3

Page A

Page B

Page C

In memory Structure (Insert Pipe)

A B C D

Page D

Insert Algorithm 2

• Applicable with UTS with Member Cluster (with and without APPEND)

• SYSTEM level : DSNZPARM: DEFAULT_INSERT_ALGORITHM

1 – use basic insert algorithm (algorithm 1)

2 – use new insert algorithm (algorithm 2) - Default for DB2 12 FL500

• Object level : CREATE TABLESPACE or ALTER TABLESPACE

INSERT ALGORITHM clause

0 – use subsystem default

1 – use basic insert algorithm (DB2 11 behavior)

2 – use advanced insert algorithm

>4 GB Active Log Data Sets

• Active log data set size impacts frequency of log switch and speed of recovery

• DB2 12 maximum size of active log data set is 768 GB

• SMS data class with extended addressability (EA) set to YES

• Customers may opt for size less than 768 GB

• Size of archive log data sets also a factor in choosing active log data set size

• >4 GB active logs supported After New Function Activation (FL500)

• Adding new active log data set > 4 GB can be done by

• DSNJU003 standalone utility or

• -SET LOG NEWLOG

• > 4GB and <= 4GB logs can be mixed

DB2 Latch Reduction

• DB2 internal latch contention relief • LC14 buffer manager latch

• LC19 log latch

• LC23 page latch timer

• LC23 page unlatch

• LC24 EDM latch

0

20000

40000

60000

80000

100000

120000

2 10 100 180

Log

latc

h c

on

ten

tio

n p

er

seco

nd

Number of Threads

Log Latch Reduction

DB2 11 DB2 12

• Simulated stock exchange transactions

• Utilizing DB2 12 scalability features : New insert algorithm and >4GB log, and log latch reduction

• All done with single z13 box

• 12 way DB2 data sharing with 4 way sysplex

• 2 zLinux driving 1200 clients concurrently

• 2000MByte active log I/O per second

• 11.7 million per second from 12 members without index defined

• 5.3 million per second from 12 members with a partition index defined

41

11.7 Million Inserts Per Second

DB2DB2

DB2

DB2DB2

DB2

DB2DB2

DB2

DB2DB2

DB2

Java clients On zLinux

z13 4 LPAR * 18 CPs512 GB Memory

Data base and Log DS8870 * 7

z13 2 ICF * 10 CPs, 200GB

Coupling Facility

z13 2 zLinux * 18 CPs, 40GB

Benchmark Environment

Addressing Customers Pain Points

Buffer pool simulation

Reduce the impact of long running unit of work in data sharing

Minimize the impact of Unload & Runstats in data sharing

A new keyword REGISTER NO/YES is added to UNLOAD / RUNSTATS

Peer Recovery support

Fast DRDA Load

DRDA Global Transaction support for data sharing

IMS connection Pooling

Possible Resource Planning with Default System Parameters Parameters • Increase in real storage requirement - up to 30% of DB2 11

Index in memory • Possibly up to 20% of buffer pools allocated, so far it has not reched

Insert algorithm 2 • 37KB per partition for each member • Additional buffers may be necessary (up to 2000 buffers per objects) depending on

the disk I/O performance EDM pool management change

• Up to 20% of specified EDM pool size Default zparm size increase

• EDM_SKELETON_POOL 10M->50M SCA minor usage increase - Data sharing lock avoidance improvement

• Additional 16KB * (N+1) to SCA for new Global CLSN support, where N = number of members to store commit LSN

• Increase in log record size • Additional 20-28 bytes at migration of DB2 12

DB2 12 Key Features - SQL Query Performance SQ

L Pe

rfo

rman

ce

• Access Path Stability • Support for dynamic SQL plan stability • Usability for static plan stability • Automated update of stats profile by optimizer

• Query Performance Improvement• Union all and outer join enhancements • Runtime improvements

• UDF caching • Sort improvement • Adaptive Index selection

Query Workload Results

0 10 20 30 40 50 60 70 80 90 100

Biday long

SAP BW

TPCD

TPCH-SQLPL

Customer 1

Customer 4

TPCH NPI D=Any

TPCH NPI D=1

Customer 2

Biday short

SAP SFIN

Customer 3cl

WAS portal (selective)

SAP SFIORI

DB2 12 CPU doff (%) from DB2 11 NFM

DB2 12 Beta Program

DB2 12 Cost Savings

The biggest benefit with DB2 12 comes with Index in memory optimization (Fast Traverse Block) which provides incredible costs savings with lower CPU consumption for OLTP – nearly 9-10% after Rebind. This should bring down our mainframe operating costs.

Jacek SurmaDII,Zespół Systemów Mainframe

DB2 12 - The #1 Enterprise Server for Mission Critical Data !

We are really excited about Performance Enhancements in DB2 12 especially advance "in-memory" (Fast Traversal Blocks FTB) capabilities. During testing we have seen up to 5 % CPU reduction and this clearly relates to enormous potential cost savings and positions DB212 as a leader in Enterprise Database market.

İbrahim ParlakIT ManagerGaranti Bank

DB2 12 Availability & Security

We are very pleased with many of the new DB2 12 features, especially with Transfer Ownership and Pending Alter Column feature this give our Enterprise higher availability and securitywhich are “critical” in the banking industry.

Jacek SurmaDII,Zespół Systemów Mainframe

DB2 12 – Exciting new capabilities

We love the "agile partition technology" that DB2 12 offers. This feature makes it easier for ITERGO to address "hot spots" where "new data" is inserted. This is particularly important when enterprises are looking for scale, speed and reduced costs .

Walter JanissenChief Architect ITERGO

DB2 12 rules the API Economy

The RESTful API is yet another way where DB2 is at the leading edge – and again cementing DB2’s and the mainframes position as a full capable server in the IT infrastructure of today. Using these REST-services Mobile applications can both be built faster and run faster !

Frank PetersenChief Architect BankData

DB2 12 –Offering Advanced “in-memory” technology

We are looking forward to exploiting the advanced "in-memory" technology that DB2 12 offers (Index Fast Traverse Block) this gives us an opportunity to reduce CPU resource consumption and performance cost by using more real memory. It is very cost effective trade off for enterprises like us that run DB2 12 on z13 machines. During testing we have seen up to 23% CPU reductions in "specific" test cases.

Henrik Henriksen

DB2 DBA Danske Bank

More than 50 Clients and Vendors

Very large DB2 users

“V11 migration was good (an improvement over V10) but V12 is even better!”

– Large German Insurance Company

Akiko Hoshikawa IBM Silicon Valley Labakiko@us.ibm.com

Notices and

disclaimers

Copyright © 2016 by International Business Machines Corporation (IBM). No part of this document may be reproduced or transmitted in any form without written permission from IBM.

U.S. Government Users Restricted Rights - Use, duplication or disclosure restricted by GSA ADP Schedule Contract with IBM.

Information in these presentations (including information relating to products that have not yet been announced by IBM) has beenreviewed for accuracy as of the date of initial publication and could include unintentional technical or typographical errors. IBM shall have no responsibility to update this information. THIS DOCUMENT IS DISTRIBUTED "AS IS" WITHOUT ANY WARRANTY, EITHER EXPRESS OR IMPLIED. IN NO EVENT SHALL IBM BE LIABLE FOR ANY DAMAGE ARISING FROM THE USE OF THIS INFORMATION, INCLUDING BUT NOT LIMITED TO, LOSS OF DATA, BUSINESS INTERRUPTION, LOSS OF PROFIT OR LOSS OF OPPORTUNITY. IBM products and services are warranted according to the terms and conditions of the agreements under which they are provided.

IBM products are manufactured from new parts or new and used parts. In some cases, a product may not be new and may have been previously installed. Regardless, our warranty terms apply.”

Any statements regarding IBM's future direction, intent or product plans are subject to change or withdrawal without notice.

Performance data contained herein was generally obtained in a controlled, isolated environments. Customer examples are presented as illustrations of how those customers have used IBM products and the results they may have achieved. Actual performance, cost, savings or other results in other operating environments may vary.

References in this document to IBM products, programs, or services does not imply that IBM intends to make such products, programs or services available in all countries in which IBM operates or does business.

Workshops, sessions and associated materials may have been prepared by independent session speakers, and do not necessarily reflect the views of IBM. All materials and discussions are provided for informational purposes only, and are neither intended to, nor shall constitute legal or other guidance or advice to any individual participant or their specific situation.

It is the customer’s responsibility to insure its own compliance with legal requirements and to obtain advice of competent legal counsel as to the identification and interpretation of any relevant laws and regulatory requirements that may affect the customer’s business and any actions the customer may need to take to comply with such laws. IBM does not provide legal advice or representor warrant that its services or products will ensure that the customer is in compliance with any law.

55 12/13/2016World of Watson 2016

Notices and

disclaimers

continued

Information concerning non-IBM products was obtained from the suppliers of those products, their published announcements or other publicly available sources. IBM has not tested those products in connection with this publication and cannot confirm the accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on the capabilities of non-IBM products should be addressed to the suppliers of those products. IBM does not warrant the quality of any third-party products, or the ability of any such third-party products to interoperate with IBM’s products. IBM EXPRESSLY DISCLAIMS ALL WARRANTIES, EXPRESSED OR IMPLIED, INCLUDING BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.

The provision of the information contained herein is not intended to, and does not, grant any right or license under any IBM patents, copyrights, trademarks or other intellectual property right.

IBM, the IBM logo, ibm.com, Aspera®, Bluemix, Blueworks Live, CICS, Clearcase, Cognos®, DOORS®, Emptoris®, Enterprise Document Management System™, FASP®, FileNet®, Global Business Services ®, Global Technology Services ®, IBM ExperienceOne™, IBM SmartCloud®, IBM Social Business®, Information on Demand, ILOG, Maximo®, MQIntegrator®, MQSeries®, Netcool®, OMEGAMON, OpenPower, PureAnalytics™, PureApplication®, pureCluster™, PureCoverage®, PureData®, PureExperience®, PureFlex®, pureQuery®, pureScale®, PureSystems®, QRadar®, Rational®, Rhapsody®, Smarter Commerce®, SoDA, SPSS, Sterling Commerce®, StoredIQ, Tealeaf®, Tivoli®, Trusteer®, Unica®, urban{code}®, Watson, WebSphere®, Worklight®, X-Force® and System z® Z/OS, are trademarks of International Business Machines Corporation, registered in many jurisdictions worldwide. Other product and service names might be trademarks of IBM or other companies. A current list of IBM trademarks is available on the Web at "Copyright and trademark information" at: www.ibm.com/legal/copytrade.shtml.

56 12/13/2016World of Watson 2016