DB2 for z/OS Distributed Access –

Best Practices and Updates

Adrian Burke
IBM

March 13, 2012

Session 10996
2
 Agenda

• Sysplex Workload
Balancing
• Parameters
• Aliases
• Cancel Thread
• Profiling
• WLM

Adrian Burke
DB2 SWAT Team
agburke@us.ibm.com
3
 DB2 Connect Configurations
DB2 Connect
Gateway Server

DRDA
DR
DA
1
DA
D R 2
Data

HTTP DRDA
IBM System z / i
3
Application
server DA
4
D R

4
 Why Would I use DB2 Connect Server
• Two Phase Commit
• Many transaction managers still require a DB2 Connect Server
for two phase commit if they use a dual transport model
• For example, Tuxedo, Encina
• DB2 and MS DTC are two transaction managers that do not
require a DB2 Connect Server
• Federation
• Homogeneous Federation is possible with DB2 Connect Server
• Use of nicknames to other DB2 and Informix data servers
• Licensing
• DB2 Connect Enterprise Edition with concurrent user licensing
requires DB2 Connect Server configurations only
• Ensure you have license for MSUs and Host (based on # of
subsystems or D.S. group
5
 Sysplex Workload Balancing

• The Sysplex Distributor

• Allows you to transparently establish an initial connection to
an available member of the data sharing group
• However, the workload balancing is network based so should
just be for establishing the initial connection to the group
• The driver/connect server uses the WLM-provided
information, together with DB2 Connect’s sysplex support,
to do the actual workload balancing of the different
connections

6
So what are the key points to Pooling and Concentrator

• Both can support Data Sharing

• Both can support VIPA and Sysplex Distributor
• Connection Pooling reuses threads at connection time
• Have to drop the connection to be routed to another DB2
• Connection Concentrator reuses threads at commit time
• Restrictions:
• Only supports SSL for outbound connections
• If you declare global temporary tables, they must be closed
explicitly at transaction or branch boundary
• Dynamic prepare requests from embedded dynamic SQL
applications will be rejected.
• See DB2 Connect user guide for details on other restrictions
• SAP does not support Connection Concentrator
• Observations:
• Most customers use Connection Pooling
• Thoroughly test usage of Connection Concentrator – often times
applications are using some of the above restrictions which do not
show up until in production

7
Sysplex Workload Balancing
• Sysplex Workload Balancing was supported by the JCC type 4
since V8 and by non-Java since 9.5 FP3
• What level driver came thru the maintenance stream…
• https://www-
304.ibm.com/support/docview.wss?q1=pk77599&rs=0&uid=swg21428742&
cs=utf-8&lang=en&loc=en_US&cc=us

JVM DB2 Universal Sysplex

Driver Type4 Distributor DB2
Thread
Logical z/OS
Connection
1
1
Transport
Logical 1
Thread
Connection
2
2
Transport
Logical 2
Thread Connection DB2
3 3
Concentration z/OS
disconnect
at commit/rollback
Pooling

8
 SysplexWLB and Failover
• Java and non-Java IBM Data Server drivers can exploit the
sysplex as follows:
• They can connect applications to a DB2 data sharing group as
though it were a single database server, and spread the
workload among the different members, based on server lists
dynamically provided by WLM (connection pooling)
• They can recognize when a member of a DB2 data sharing
group fails and can automatically route new connections to
other members. This is done after commit scope of application,
otherwise connection would simply hang. (connection
concentration)

9
DB2 Inactive Thread - Tuning Considerations

Application DB2 DB2 zparm:

Server Connect/JCC:
•MAXDBAT
Connection •max Agents/
•CONDBAT
Pool: Transports
•IDTHTOIN
•unused •Connections
•DB2_MAX_
timeout
CLIENT_
•max CONNRETRIES)
connection

10
Thread Monitoring:
-DIS DDF DETAIL real time information on DBAT and CONDBAT metrics
-------------------------------------------

DSNL080I +APDB DSNLTDDF DISPLAY DDF REPORT FOLLOWS:

DSNL081I STATUS=STARTD
DSNL082I LOCATION LUNAME GENERICLU
DSNL083I APDB DHHR.APDB
How many threads are -NONE
DSNL084I IPADDR currently
TCPPORT
doingRESPORT
work
DSNL085I 10.116.2.5 5251 5252 What you think you can
DSNL086I SQL DOMAIN=share.org handle.
DSNL086I RESYNC DOMAIN=share.org
DSNL090I DT=I CONDBAT= 10000 MDBAT= 200
DSNL092I ADBAT= 0 QUEDBAT= 0 INADBAT= 0 CONQUED= 0
DSNL093I DSCDBAT= 0 INACONN= 5
DSNL099I DSNLTDDF DISPLAY DDF REPORT COMPLETE
*** You hit max DBAT and
How many threads are some are waiting
lounging in the pool
Type 2 inactive thread/
inactive connection
11
 Configuration Best Practices (non-JAVA)

• enableWLB – True
• Enables WLB and seamless failover.
• Transports allocated based on app connection
• Default max of 1000 transports per member
• ConnectionTimeout – Network. Based on Application QoS
• Connect will fail when expired
• CommandTimeout - Network (.NET 30 seconds). Based on Application
QoS.
• Triggers interrupt when expired.
• maxAcrRetries – 1 time thru server list
• Number of times to retry the connection to the alternate server.
• AcrRetryInterval – 1 second
• Number of seconds to wait between retries.

12
 Configuration Best Practices (non-JAVA)
• maxTransportIdleTime - 60 seconds (< than IDTHTOIN)
• Time an unused transport hangs around before being closed.
• maxRefreshInterval – 10 seconds
• Time between updates of WLM weights with server.
• maxTransportWaitTime – 1 second
• Max time an app will wait for a transport to become available.
• memberConnectTimeout - 1 second
• Timeout to be used when communicating with a member in the server
list.
• tcpipKeepAlive – 15 seconds (< than TCPKPLV)
• Timeout for all other communication requests
• interruptprocessingmode – 2
• Controls interrupt behavior. Value=2 drops the connection when
interrupt is triggered.

13
 Configuration Best Practices…
• Cursor Hold – we do not want result sets persist across commits so…
• JAVA – setResultHoldability =2
• Non-JAVA – CursorHold=0

• Implicit Close – we want cursors to be automatically closed after rows

are fetched so…
• JAVA – queryCloseImplicit
• Non-JAVA – CursorTypes

• Auto Commit – we would like the application to intelligently moderate

the commit scope through the persistence layer or manually within the
application
• JAVA – setAutoCommit(false)
• Non-JAVA – Autocommit configuration keyword

14
 Key DB2 DDF Parameters
Default – as of
Parameter Possible values V8* Description

DDF NO / AUTO / COMMAND No DDF Startup

CMTSTAT ACTIVE / INACTIVE INACTIVE* Thread Pooling - Pool inactive threads

Max users - allied (local) threads RRSAF or CAF

CTHREAD 1-2,000 200* (CICS, IMS, TSO, Batch attach, SPUFI, Classic QMF, etc)

MAXDBAT 0-1,999 200* Max remote active DDF Threads - DBM1 Address Space

CONDBAT 0-15,000 10,000* Max remote connections - DDF Address Space

Max inactive DBATs, these are used for private protocol.

MAXTYPE1 0-CONDBAT 0 DRDA uses inactive connections.
These were
mentioned on
previous slides
Approximate time, in seconds that an inactive/unused DBAT can remain idle in
POOLINAC 0-9,999 120 the pool before it is terminated. DBAT deleted after being used 200 times also.

The IDLE THREAD TIMEOUT (IDTHTOIN)

parameter specifies the time (in seconds) that an active
DBAT can remain idle before it is canceled. – Should set
IDTHTOIN 0-9,999 120* a “little” higher than TCPKPALV
ENABLE / DISABLE / 1-
TCPKPALV 65534 120* TCP/IP keep alive (Goes hand-in-hand with IDTHTOIN)

CONTSTOR YES / NO No Periodically ″contract″ each thread’s working storage area.

Use storage management algorithms that minimize the

MINSTOR YES / NO No amount of working storage consumed by individual threads.

15
 WLM Weight calculation.
There are no configuration parameters related to enabling sysplex WLB on DB2
for z/OS server. When DDF is started, it registers itself to WLM, unless
MAXDBAT is set to 0 or DDF is stopped. In either of these exception cases,
that member’s DDF does not appear in the server list. The drivers use the
updated server list on each transaction to distribute workload to DB2 data
sharing members.

There are several factors WLM will take into account when creating and updating
weights.
1. Displaceable capacity of systems (CPU).
2. Enclave service class achievement (Performance Index, or PI)

WLM goal should be attainable when system is not under stress, which would
result in a PI < 1.
3. Enclave service class queuing.
4. DB2 for z/OS Health

DB2 will report its health factor of 0 to 100 to WLM based on the current
storage consumption within the DBM1 and DIST address spaces.

16
 Subsetting configuration
DB2 LOCATION

DB2 LOCATION ALIAS

DB2A z/OS-1
z/OS-2 z/OS-3
DB2B DB2C
Vx,446 Vx,447 2
V1,446 V1,447 Vx, 446 Vx,447
Dispatch
V1,5001 connection V2, 446 V2,447 Vx, 446
to DB2B V3, 446
V2, 5002
V3, 5003
SD: Vx

1 3
Initial connection SRVLST returned (V1:W1, V2:W2)
to DB2A or DB2B
using Vx,447

4
4 Workload balancing to
DB2B using V2,447

1 First connection goes through Sysplex Distributor (DVIPA)

2 First based on IEAFSSNxx order in SYS1.PARMLIB

3 A list of available DB2 subsystems and their respective weightings are returned

4 Based on weighting following connections and threads are routed

17
 Types of location aliases

• Static location aliases

You can use DSNJU003 (change log inventory) utility to define and
modify as many as 8 static location aliases. Changes to these aliases
require you to stop both DDF and DB2, thus requiring an outage.

• Dynamic location aliases (New in v10)

You can use the MODIFY DDF command with the ALIAS option to
define and manage as many as 40 location aliases dynamically. You can
start, stop, cancel, modify, and delete dynamic location aliases without
stopping either DDF or DB2. These aliases cannot be defined or
managed by the DSNJU003 utility, and the DSNJU004 utility does not
print any information about them. You have to use the DISPLAY DDF
command to find information about these aliases. Before you can define
dynamic location aliases, DB2 must be started, but DDF may or may not
be started.

18
Dynamic Location Aliases

Business Challenge : Need the ability to manage poorly

performing app servers without taking a DB2 server outage.
Solution: You can associate a location alias with an app server
and use the –MODIFY DDF command to dynamically exclude
poorly behaved applications from sysplexWLB (change PORT)
Business Challenge : Stopping DDF causes all the app servers
connected to a member to failover. DBAs are looking for a
more granular failover where only certain app servers fail over
while others don’t.
Solution: You can associate a location alias with an app server
and use the –MODIFY DDF command to dynamically stop the
alias which only causes the app servers connected to that alias
to failover, without affecting others.

19
Dynamic Location Aliases
Business Challenge : Block incoming traffic from certain app
servers without disrupting outbound traffic. Setting MAXDBAT=0
blocks all inbound traffic.
Solution: You can associate a location alias with an app server and
use the –MODIFY DDF command to dynamically stop the alias
on a member which blocks incoming traffic from only those app
servers that are connected to that alias without affecting other
inbound or any outbound traffic.
Business Challenge : The default IP addresses provided to the
client drivers may not always be usable to establish connections
for sysplexWLB or failover. So, DBAs want to control the IP
addresses provided to the client drivers.
Solution : You can use the –MODIFY DDF command with the
ALIAS keyword to dynamically specify the IP address that will be
returned in the server list when client connects to the specified
alias.

20
Dynamic Location Aliases - Continued
Examples :
-MODIFY DDF ALIAS(alias1) ADD
Alias1 is created and is stopped by default.
-MODIFY DDF ALIAS(alias1) PORT(9000)
Alias1 is associated with port 9000.
-MODIFY DDF ALIAS(alias1) IPv4(2.2.2.2)
Alias1 is associated with IP address 2.2.2.2
-MODIFY DDF ALIAS(alias1) START
DDF will accept requests for alias1 on port 9000.
When a client connects to alias1, IP address 2.2.2.2
is returned in the server list.
-MODIFY DDF ALIAS(alias1) STOP
Alias1 is stopped and will not accept new requests.
Existing requests will be allowed to complete.

21
 Displaying location alias-specific
information
-DISPLAY DDF ALIAS(alias1 ) DETAIL
DSNL080I @ DSNLTDDF DISPLAY DDF (alias1) REPORT FOLLOWS:
DSNL087I ALIAS PORT SECPORT STATUS
DSNL088I ALIAS1 9000 5005 STARTD
DSNL089I MEMBER IPADDR=::2.2.2.2
DSNL089I MEMBER IPADDR=2002:91E:610::1
DSNL096I ADBAT= 100 CONQUED= 1000 TCONS= 1000
DSNL100I LOCATION SERVER LIST:
DSNL101I WT IPADDR IPADDR
DSNL102I 32 ::2.2.2.2 2002:91E:610::1
DSNL102I 32 ::1.2.3.4
DSNL099I DSNLTDDF DISPLAY DDF REPORT COMPLETE

ADBAT : The number of active data base access threads (DBATs) that are currently
processing requests on behalf of the specified alias.
CONQUED : The number of connection requests that are currently queued and waiting to be
serviced on behalf of the specified alias.
TCONS : The total number of remote connections that are currently associated with the
alias.

22
 Online CDB (Communications DataBase)

Business Need : TCP/IP connections

established after modifying the CDB should
automatically use the updated values without
recycling DDF.
Solution : When a CDB table used for
TCP/IP access is updated, DDF is notified
and it dynamically retrieves the updated
values to establish new connections. You can
use the –DISPLAY LOCATION command to
tell which connections are using the updated
CDB values and other connection attributes
like AES, SSL, IPSec, trusted context, XA or
sysplexWLB.

23
Cancel Thread improvements

Business Challenge : Lack of sufficient cancel

detection points in DB2 server is causing certain
queries to run indefinitely and consume
resources, even after client disconnects. DBAs
want to be able to cancel a long running query
and not be charged for wasted CPU costs that
will never be materialized to the applications.
Solution : More cancel detection points have
been added in DB2, mainly in areas like sort and
work file processing to catch more runaway
threads.

24
Cancel Thread improvements - Continued

Business Challenge : DBAs want to be able to cancel

requests suspended in the network effectively
without issuing external commands to drop the
connection. Users may not know it is suspended in
the network and end up recycling DB2 to free the
held resources.
Solution : DB2 will now programmatically drop the
connection using TCP/IP network manager calls that
allows DB2 to get control and cancel the thread.
APAR number is PM54383.

25
System Profiling

Business Challenge : With direct

connections, DBAs need greater
flexibility and control to manage (punish)
resources for specific clients according
to their needs.
Solution : You can use DB2 profile tables
to monitor connections for specific
clients. You can specify thresholds and
actions to be taken when threshold is
exceeded. You can set filters to identify
client connections to be monitored.

26
System Profiling - Continued

Filters that you can set.

• Client IP address/domain name
• Client product id
• Client application name
• Client userid
• Client workstation name
• Role name, authid
• Collection id/package name
• Location/location alias name

27
System Profiling - Continued
What can be monitored ?
1. MONITOR CONNECTIONS
• Monitor the number of concurrent inbound connections allowed.
• Is subject to the filtering on IP address or Domain name only.
2. MONITOR THREAD
• Monitor the number of DBATs that are allowed to be concurrently
active.
• Is supported for all filtering combinations.
3. MONITOR IDLE THREADS
• Monitor the amount of time, in second, that an active server thread
is to be allowed to remain idle.
• Is supported for all filtering combinations.
• You can use history tables to know what profiles and monitoring
functions are in effect.
• When a profile warning or exception occurs, IFCID 402 (stat class 4)
record is written at an interval set in the STATIME zPARM.

28
Profile enhancements
• Catalog table holds profiles
• Can limit:
• Warning only produces
messages
• Exception will take action
• Warning or exception
• DIAG level 1 – 1 message
if any threshold reached
• DIAG level 2 – 1 message
for each unique threshold
surpassed
• Exception for active thread
means 10 can be active, 10
can be in queue, but next one
in gets canceled
• DSNT771I/DSNT772I console
message, reason code 00E3050x
alerts you of warning or exception
29
Protecting DB2

Business Challenge : DBAs want to control the connection

queue depth and the wait time for a queued connection to
get a DBAT, exceeding which connection should be
terminated.
Solution : Two new zPARMs MAXCONQN (depth) and
MAXCONQW (wait time) are supported. -DISPLAY DDF
DETAIL will show the current configured values and
connections closed because limit exceeded. This info is also
recorded in stats and accounting. APAR number is
PM43293 (V9 & V10).

30
Protecting DB2 - continued

Business Challenge : WLM weights are not updated until PI index

interval of 10 seconds which causes clients to use a
stale server list to route requests. The requests may
thus land on members that have a connection count
approaching the CONDBAT limit.
Solution : DB2 will lower its health when connection count starts
approaching the CONDBAT limit. Since the health is
factored in the weight calculation, clients will be
discouraged to route here. Every 1 minute, DB2 will
check if connection count started declining and will
raise back the health to the system value. The health
reported to WLM can be displayed using –DISPLAY
DDF DETAIL. APAR number is PM43293.

31
 Location Monitoring

• 24/7 availability and bigger counters : Previously stat

counters were 4 bytes and wrapped quickly. For 24/7
availability, counters have been extended to 8 bytes.

• New IFCID 365 (class 7) : Previously, stats for all locations

were grouped under DRDA REMOTE LOCS and written to
SMF every minute. With IFCID 365, stats can be displayed by
location and written to SMF at the interval specified by
zPARM STATIME. You can provide tuning info for each app
server depending on where work is coming from.

32
 Monitoring DBAT’s WLM goals (DB2 10)
• You can use the –DISPLAY THREAD command to find out the following
WLM characteristics associated with the DBAT (separated by a colon)
• Service Class Name – Named group of work with similar goals and
resource requirements.
• Importance Level of the period - The relative importance of the
service class period goal. Only used when goal is not being met. The
value can be 1-5, or DISCRETIONARY(5), 1 being the highest
importance.
• Service class Period Number - Performance periods are available
for work that has variable resource requirements and for which your
goals change as the work uses more resources. You specify a goal,
an importance, and a duration for a performance period. You can
specify up to eight performance periods.
• Performance Index of the service class period - A calculation of
how well work is meeting its goal. For work with response time goals,
PI is the actual divided by goal.
• PI=1 (period is exactly meeting its goal)
• PI<1 (period is beating its goal)
• PI>1 (period is missing its goal)
• DISCRETIONARY = 0.8
33
 Example Service Period
Class

-Display Thread(*) Type(Active) Detail Importance

DSNV401I > DISPLAY THREAD REPORT FOLLOWS -

DSNV402I > ACTIVE THREADS - Performance
index
NAME ST A REQ ID AUTHID PLAN ASID TOKEN
TEST0001 RA * 1 CORID001 SYSADM DONSQL1 0050 6
V441-ACCOUNTING=HADERLE
V482-WLM-INFO=DDFSVCLS:1:5:1.1
V442-CRTKN=9.30.113.201.5001.C68EA0176B20
V445-USIBMSY.SYEC717A.C68EA0176B20=6 ACCESSING DATA
FOR
( 1)::FFFF:9.30.113.201
V447--INDEX SESSID A ST TIME
V448--( 1) 447:1027 W R2 1025213462763
DISPLAY ACTIVE REPORT COMPLETE
DSN9022I - DSNVDT '-DISPLAY THREAD' NORMAL COMPLETION

34
How Should I Classify Threads:
Work qualifiers are used to help identify a thread
• AI - Accounting Information* •PFG - Perform Group
• CI - Correlation Information* •PK - Package Name *
• CN - Collection Name •PKG - Package Name Group
• CT - Connection Type •PN - Plan Name
• CTG - Connection Type Group •PNG - Plan Name Group
• LU - LU Name •PR - Procedure Name
• LUG - LU Name Group •PX - Sysplex Name
• NET - Net ID •SI - Subsystem Instance
• NETG - Net ID Group •SIG - Subsystem Instance
• PC - Process Name Group
• PF - Perform • SSC - Subsystem Collection
* Popular choices • UI - Userid

• Best Practice
• Ensure all remote workload is less important than DDF address
space
• Classify transactionary work based on response time goals (2
periods), not velocity goals – AND NOT DISCRETIONARY
35
 WLM classification example
• This is an example of ill-fated goals
• DDF workload imp. 1 high velocity goal
• DB2 DIST high velocity goal as well
• Deadly embrace as DDF uses resources DB2 needs to accept new work
• DDF response time could go up by factor of 10x, because duration
is not part of the goal nor monitored by WLM with a velocity goal
• If you continuously miss your goal WLM will skip you

36
How do I set Client-info to better classify trans.? Ease of
implementing
• In data source
But static
• All applications sharing data source appear the same to
DB2
• Need source per application to change information
• Calling WLM_SET_CLIENT_INFO stored procedure
• Requires application add a call to proc and populate the
information
• pureQuery product – makes JAVA into static packages
• Having application set it
• WSConnection() method to set correlation and accounting
info
• Create a wrapper from incoming getConnection() string that
dynamically picks up program name and IDs
• Can use Hibernate or Spring class to populate their
intermediary config file
• Could use a wrapper from Websphere that uses
getConnection() and WSSubject class to pull the Requires
information out of the incoming request to populate client coding, but
info flexible

37
Example of Creating a DDF Service Class for a Specific
Application Continued . . .
• If a service class of ‘PC=AdrianBurke’, or ‘CI=agb_v9’ had been
created and the application code contained
WSConnection.CLIENT_APPLICATION_NAME, “agb_v9”, or
connectionProperties.put("clientProgramName", "agb_v9"), and
connectionProperties.put("clientApplicationInformation",
“AdrianBurke") in connection string: then the snapshot of the enclave
screen would show the following details:

38
Distributed Correlation Enhancements
Previously, the correlation token used to correlate work between the client and
server was only externalized in DB2 accounting data and not in DB2
messages making it impossible for users to correlate message related
failures to the remote client application that is involved in the failure.
In DB2 10, the THREAD-INFO description for key messages will include this
token, which will be enclosed in '<‘ and '>' characters, and contains three
components,
separated by a period:
• A 3 to 39 character IP address.
• A 1 to 8 character port address.
• A 12 character unique identifier.
Example :
DSNL027I > SERVER DISTRIBUTED AGENT WITH
LUWID=G91702F8.P853.100629180434=4
THREAD-INFO=ADMF001:mask:admf001:db2bp:*:*:*:
<9.23.2.248.38995.100629180434>
RECEIVED ABEND=04E
FOR REASON=00D3001A

39
 Optional Domain name

• Previously, we did not allow DB2 to process TCP/IP requests without

configuring a domain name, even when the IP address that domain name
maps to is always fixed.

• In v10, for users that specify fixed IP addresses in the DB2 BSDS, a
domain name is no longer required to be configured to process TCP/IP
requests.

• When a domain name is unavailable, A DSNL523I message that

contains the BSDS-specified IP address will be issued, in lieu of a
DSNL519I message to indicate that DDF is ready to accept
requests for that IP address.

40
Whats new with remote threads in DB2 10…
• High Performance DBATs is a new type of distributed thread
(some customershave seen XX% perf. Gain)
• Packages must be bound with RELEASE(DEALLOCATE) to get
reuse for same connection and -MODIFY DDF PKGREL(BNDOPT)
must also be in effect
• Since V6 they have been RELEASE(COMMIT)
• Now DBAT and client connection will remain active together
• After the Hi-Perf DBAT has been reused 200 times
• DBAT will be purged and client connection will then go inactive

• If Hi-Perf DBAT has not received new work for POOLINAC time
• DBAT will be purged and the connection will go inactive

• If # of Hi-Perf DBATs exceed 50% of MAXDBAT threshold

• DBATs will be pooled at commit and package resources copied/allocated as
RELEASE(COMMIT)

• Hi-Perf DBATs can be purged to allow DDL, BIND, and utilities to

break in
• Via -MODIFY DDF PKGREL(COMMIT)
41
 Cost of deployment for type 2 vs. type 4 drivers

• Type 4 saved CPU over 3 years ago

• Even then T4 used 23% more total
processing cycles (PTV8)
• 23% less ITR (PVT8)
• 19% increase in elapsed time (PVT8)
• Current V10 numbers show T2 saves 5%
on CP, and 17% CP+zIIP
http://www-
03.ibm.com/support/techdocs/atsmastr.n
sf/5cb5ed706d254a8186256c71006d2e0
a/94f9d53f2e526489862575c5004b50b1/
$FILE/WP101476-2%20-
%20Value%20of%20Co-
Location%20Update.pdf

42
 Extras……
• No-Charge workshop (email me)
• System z Synergy workshop focused on Websphere (LUW or
z) and DB2 for z/OS, settings, best practices, lessons learned

43
 Questions???

Adrian Burke
DB2 SWAT Team SVL Lab
agburke@us.ibm.com

• VISIT the DB2 Best Practices

• VISIT the DB2 for z/OS Exchange
• JOIN the World of DB2 for z/OS
• JOIN the DB2 for z/OS group

44