Application Tuning for CA Datacom Environments Part II

World®
’16
Application Tuning for CA
Datacom environments – Part II
Kevin Shuma, VP Product Management
CA Technologies
MFX67EB
MAINFRAME AND WORKLOAD AUTOMATION

2 © 2016 CA. ALL RIGHTS RESERVED.@CAWORLD #CAWORLD
Abstract
This is a two part session that describes the various techniques
used to tune a CA Datacom application for the best possible
performance. Included will be discussions on available
monitoring tools.
This session is primarily focused on the application developer
who is writing applications for the CA Datacom®/DB database.
Kevin Shuma
CA Technologies
VP, Product
Management

Agenda
APPLICATION TUNING WITHIN THE APPLICATION
WHAT IS THE URT AND HOW DOES IT AFFECT THE APPLICATION
GENERAL PROGRAMMING RECOMMENDATIONS
IDEAL PROGRAMMING RECOMMENDATIONS
AVAILABLE TUNING TOOLS AND RESOURCES
SUMMARY
1
2
3
4
5
6

Application Tuning within the application
§ Some application issues need to be resolved within the
application
§ In other cases, new or different programming techniques can
be used to improve the applications performance
§ This session will address ways to improve application
processing within the application program
Best practices that can be applied

What is the URT and how does it
affect the application

What is a URT?
§ Much more than just a list of tables accessed by the program
– Interface code (batch non-Ideal)
– Processing options
– Tables list and Sequential Read Ahead options
– GBMAXR/GETBLK/SEQBUFS
– GSETL/GETIT invoked automatically by CA Ideal FOR constructs under
the right circumstances
– Not invoked automatically for CA MetaCOBOL+
Two-part webcast is available on the subject of URTs

Sequential Read Ahead
Simplified view
Half of SEQBUFS
buffers (1-16) filled
NOW
Second half of SEQBUFS
buffers (17-32) filled
while first half are
being processed
Application Region MUF Region
GSETL/GETIT
SEQBUFS=32
GETBLK=61440
GBMAXR=255
Block of rows returned
subject to GBMAXR limit
(255 is “unlimited”)

Sequential Read Ahead (cont’d)
§ Pros
– Significantly faster batch sequential processing
– Reduced elapsed time due to MUF read-ahead
– Reduced number of MUF to application communications
§ Especially important when using XCF
– Possible reduced IOs due to very fast buffer re-use
Simplified view

Sequential Read Ahead (cont’d)
§ Cons
– Significant possibility for application program to misuse the
functionality
§ Getting lots of data rows that are never actually read in the program
§ Typical when not reading the key range from beginning to end
§ For non-Ideal, be sure to specify the key range end value
– Getting lots of data rows and holding them under exclusive control
§ If URT has UPDATE=YES all rows passed in Block GETxx buffer are locked
§ Can be minimized by using smaller GBMAXR and/or GETBLK (e.g.
GBMAXR=10)
Simplified view

URT Priority
§ MUF BREAK parameter will influence effectiveness
– BREAK 500 suitable at most sites
§ What to specify for PRTY= in URTs:
– Hot CICS AORs and Hot Trusted Server: 12
– Normal CICS AORs and Trusted Server: 10
– Hot batch (Time-sensitive or Critical Path): 8 or 9
– Normal batch: 7 (CA Default)
– Dataquery CICS AORs and Ad-hoc Server: 5
– Dataquery batch, DBSQLPR and Data Reporter: 3
Does it still really matter? In a busy MUF, Yes!

General programming recommendations

General Programming Recommendations
§ Record-at-a-Time (RAAT)
§ Set-at-a-Time (SAAT or “CBS”)
§ SQL
§ Can mix and match all three access methods within a single
program
One DBMS, three APIs available to 3GLs (COBOL, PL/I, etc.)

Record-at-a-Time (RAAT)
§ Most commonly used API for legacy applications
– Not available directly to programmers in CA Ideal
– Used by CA MetaCOBOL+ LOCATE/READ statements
– Similar to VSAM programming API
§ All processing selections predetermined (table, keys,
elements)
– Changes to keys or elements require program changes
§ Very fast for direct random keyed access (LOCxx/REDxx)
– But each row is returned individually – communications overhead

Record-at-a-Time (RAAT) (cont’d)
§ Sequential read ahead (GSETx/GETxx)
– Batch only
– Very fast if trawling through large key ranges
– Tune GBMAXR/GETBLK/SEQBUFS in URT
– “New” feature (r12) to specify end-of-key-range on GSETL
§ Often rows are returned that are later rejected by the
application in non-zIIP processing

Set-at-a-Time (SAAT or “CBS”)
§ API built for use with 4GL applications
– Similar to SQL programming API (without joins)
– CA Ideal and CA MetaCOBOL+ “FOR” constructs
§ Call DBNTRY….SELFR/SELNR/SELSM/SELPR
§ Dataviews can only reference a vertical subset of a single table
§ Can select on non-key columns, and order rows
– Reducing communications overhead and performing the “sorting” on zIIP
– Review old programs with IF tests and internal sorts

Set-at-a-Time (SAAT or “CBS”) (cont’d)
§ CA MetaCOBOL+ does NOT convert SRA candidates
to GSETL/GETIT
§ CA MetaCOBOL+ (or CALL) does allow SKIP processing
within the set
– Change direction or jump to top/bottom of set, or relative position
– Great for simple (single-table) scrolling applications

§ Tune using CBS trace – “$$$” in chars 1-3 of UIB
§ Manual joins (nested FOR) often means rows are returned
that are later rejected by the application in non-zIIP
processing

§ The temporary index
– CBS examines the request made and determines how to access the data
– A temporary index is built if
§ Mixed ascending/descending ORDER BY
§ ORDER/GROUP BY a non-key field
§ Signed non-positive numeric fields possible in key
§ CBS chose a better key to satisfy the WHERE clause
– Sometimes a temporary index is NOT a bad thing.
§ Unnecessary ones should be eliminated

CBS - Building a temporary index
Mixed ORDER BY with both ascending and descending
FOR FIRST PRICE-LIST-R001
WHERE EFCTV-DATE = W.EFCTV-DATE
AND FGI-ID = W.FGI-ID
AND PRICE-CLS-ID = W.PRICE-CLS-ID
AND ALWN-BRCKT-CDE GT $SPACES
ORDERED BY EFCTV-DATE FGI-ID PRICE-CLS-ID
DESCENDING ALWN-BRCKT-CDE
FOR FIRST PRICE-LIST-R001
WHERE EFCTV-DATE = W.EFCTV-DATE
AND FGI-ID = W.FGI-ID
AND PRICE-CLS-ID = W.PRICE-CLS-ID
AND ALWN-BRCKT-CDE GT $SPACES
ORDERED BY DESCENDING EFCTV-DATE
FGI-ID
PRICE-CLS-ID
ALWN-BRCKT-CDE
Before After
All the “ascending” columns in the ORDER BY are specified as
equalities in the WHERE clause, so the ORDER BY can be changed
without impacting the program

CBS - Building a temporary index (cont’d)
§ SQL CREATE TABLE statements always define numeric columns
as signed
– DBA can change to unsigned afterwards
§ If the column only ever contains positive values
– DBA should set TYPE-NUMERIC=P attribute
§ If the column does contain negative values
– DBA should set DATA-SENSITIVE attribute
Signed numeric columns in key

CBS - Determining the best key to use for CBS
§ CBS dynamically chooses the best key to use based on
information in the where and order by clauses
§ Be careful when trying to second-guess CBS
– Run a CBS trace to check that your assumption was correct
– With two similar candidate keys, Statistical Population Counting can
change the key selection at runtime
§ In the following example
– The ORDER BY matches the native key
– But a better key matches the WHERE clause

CBS - Determining the best key to use for CBS (cont’d)
Order by determines the key – but is not the best choice
FOR FIRST SALES-EST-PD
WHERE MARKET-CODE EQ W.MARKET-CODE
AND SOURCE-CODE EQ $SPACES
AND CURRENT-WEEK EQ W-WEEK
AND CPCK-ITEM-CODE EQ $SPACES
AND SETUP-WEEK-10 GE SETUP-DATE
ORDERED BY MARKET-CODE
SOURCE-CODE
CURRENT-WEEK
SETUP-WEEK-10
FOR FIRST SALES-EST-PD
WHERE MARKET-CODE EQ W.MARKET-CODE
AND SOURCE-CODE EQ $SPACES
AND CURRENT-WEEK EQ W-WEEK
AND CPCK-ITEM-CODE EQ $SPACES
AND SETUP-WEEK-10 GE SETUP-DATE
ORDERED BY MARKET-CODE
CURRENT-WEEK
SOURCE-CODE :Moved
CPCK-ITEM-CODE :Inserted
SETUP-WEEK-10
Before After
Statistical Population Counting and a Temporary Index Now a direct selection

CBS - More WHERE clause tips
§ Avoid NOT operator, if feasible
– WHERE ACTIVE = “Y”… rather than…
– WHERE ACTIVE <> “N”
§ Specify an alternative to “OR”-ing values for a field - use a
range, if feasible
– WHERE QTY = 3 OR 4 OR 5 …becomes…
– WHERE QTY >= 3 AND QTY <= 5
– Convert to IF statement ONLY in special circumstances

SQL
§ Agile/DevOps - Quick to develop/maintain
§ All joins/aggregation/formatting done in MUF
– Available for zIIP processing
– Keeps the application small/simple/efficient
§ Minimizes communications overhead
– SQL pre-fetch works for CICS requests too
§ If there is syntax/functionality you need…
– Submit your enhancement “Ideas”
Industry standard syntax

SQL (cont’d)
§ SQL Constraints can be used to implement business rules
– Runs in MUF (on zIIP)
– Reduces need for program “checks”
§ Don’t allow a delete of a customer that still has an order
§ Only allow type “A,B,C”
§ Utilize CASCADE to trigger full deletes
§ SQL CACHE can be implemented to improve repeated
dynamic queries
Industry standard syntax

SQL tuning tips
§ IN vs. OR
– WHERE CAT_TYPE IN(’05’,’07’)
– WHERE CAT_TYPE = ’05’ OR CAT_TYPE = ’07’
– Usually OR is better
§ OR vs. BETWEEN
– WHERE CAT_TYPE IN(’05’,’06’,’07’)
– WHERE CAT_TYPE = ’05’ OR CAT_TYPE = ’06’ OR CAT_TYPE = ’07’
– WHERE CAT_TYPE BETWEEN ’05’ AND ‘07’
– Usually BETWEEN is better
Simple where clause predicate tips

SQL tuning tips (cont’d)
§ Index cardinality stats and Row Counts
– RETIX DBID=100,KEYNAME=*SETR
– REPORT AREA=IXX,DBID=100,TYPE=G,UPDATE=YES
– Don't forget UAT environments
– Especially if refreshed using DASD utilities
§ SYSMSG enhancements
– Much easier to understand
– Predicate messages no longer “lost” in PXX
– SELECT * FROM SYSADM.SYSMSG WHERE PLANNAME = 'KEVIN';
SQL query optimization housekeeping

COBOL tuning tips
§ COBOL STRING/UNSTRING with pointer is very efficient
§ Caching using internal arrays
– Good or bad?
– Direct access can be good
– Searching is CPU-intensive
§ Order the EVALUATE predicates by most-used -> Least-used

COBOL tuning tips (cont’d)
§ CBL OPTIMIZE – Dual compile?
§ Enterprise COBOL 5.x and beyond - the biggest step change
since COBOL II
– Architecture exploitation (ARCH)
– Choose the highest level supported by your hardware
– Including your DR site

Ideal programming
recommendations

Ideal programming gotchas - $SUBSTR
§ $SUBSTR is one of the most CPU-intensive functions of CA
Ideal
– Never code a $SUBSTR with fixed values for both START= and LEN=
– Break the target field into a group field and reference the data directly
as a simple field
§ Don’t repeat the same $SUBSTR function unnecessarily
– SET the result into a simple field and then reference that field in
subsequent statements
Ideal is a powerful language but some processes can be expensive to use

Ideal programming gotchas - $SUBSTR (cont’d)
Level Field Name T I Ch/Dg
----- ------------------- - - -----
1 SRA-DATA X 80
SELECT FIRST ACTION
WHEN $SUBSTR(SRA-DATA,START=70,LENGTH=3) = 'ELM'
SET ELM = TRUE
WHEN $SUBSTR(SRA-DATA,START=70,LENGTH=3) = 'DVW'
SET DVW = TRUE
WHEN $SUBSTR(SRA-DATA,START=70,LENGTH=3) = 'PGM'
SET PGM = TRUE
ENDSEL
Level Field Name T I Ch/Dg
----- ------------------- - - -----
1 SRA-DATA
2 FILLER-1 X 69
2 SRA-ENT-TYP X 3
2 FILLER-2 X 8
SELECT FIRST ACTION
WHEN SRA-ENT-TYP = 'ELM'
SET ELM = TRUE
WHEN SRA-ENT-TYP = 'DVW'
SET DVW = TRUE
WHEN SRA-ENT-TYP = 'PGM'
SET PGM = TRUE
ENDSEL
Before After

Ideal programming gotchas - $STRING
§ $STRING is one of the most CPU-intensive functions of CA
Ideal
– Data is copied to a temporary work area, appended and then copied to
the target area
– Where appropriate, use the new Ideal 15.0 DOCUMENT feature with
variable-length fields to accomplish the task
§ Don’t repeat the same $STRING function unnecessarily
– Set the result into a simple field and then reference that field in
subsequent statements

Ideal programming gotchas - $STRING (cont’d)
SET W.DUPLICATE = FALSE
LOOP
VARYING W.SUB2 FROM 1 BY 1
UNTIL W.DUPLICATE
OR W.SUB2 = 640
IF DUP-PGM-DVW(W.SUB2) =
$STRING(RPT.PROGRAM,
RPT.STATUS,
RPT.DATAVIEW)
SET W.DUPLICATE = TRUE
ENDIF
ENDLOOP
Before

Ideal programming gotchas - $STRING
§ While we are here what else can we improve?
– Avoid looping through the whole array
§ In this case, all 640 entries are searched if the incoming item is NOT
a duplicate
§ Store the “maximum entry#” value when the array is built and use
that value to terminate the loop
– Avoid re-testing the “terminator” value on each iteration of the loop

LOOP
UNTIL W.DUPLICATE
OR W.SUB2 = 640
RPT.STATUS,
RPT.DATAVIEW)
ENDIF
ENDLOOP
Before
SET NEW-PGM-DVW = $STRING(RPT.PROGRAM,
RPT.STATUS,RPT.DATAVIEW)
<<SEARCH-PGM-DVW>>
LOOP
VARYING W.SUB2 FROM 1 BY 1 UP THRU W.DVW-MAX
IF DUP-PGM-DVW(W.SUB2) = NEW-PGM-DVW
QUIT SEARCH-PGM-DVW
ENDIF
ENDLOOP
After

LOOP
UNTIL W.DUPLICATE
OR W.SUB2 = 640
RPT.STATUS,
RPT.DATAVIEW)
ENDIF
ENDLOOP
Before
<<SEARCH-PGM-DVW>>
LOOP
IF DUP-PGM-DVW(W.SUB2) = NEW-PGM-DVW
QUIT SEARCH-PGM-DVW
ENDIF
ENDLOOP
After
LOOP
UNTIL DUP-PGM-DVW(W.SUB2) = NEW-PGM-DVW
ENDLOOP
IF W.SUB2 > W.SUB
ELSE
ENDIF
After
If you really hate QUIT

Ideal programming gotchas – Internal sorting
§ AVOID unnecessary internal sorts
– Use SQL to join/aggregate/sort data into required sequence
§ Move work to MUF where zIIP CPU can be used
– Use the sorted report facility of CA Ideal instead, if appropriate
– If internal sort is the only option, avoid repeatedly calculating the index
subscript
§ Calculate the “subscript – 1” once and use that in all subsequent statements
§ Use group field to move whole array “row” at once

If internal sort is the only option, avoid repeatedly
calculating the index subscript
LOOP VARYING W.SUB FROM W.ENTRIES BY -1 DOWN THRU 1
IF W.CURR-KEY < W.PREV-KEY
SET WT.PROD-CD = W.PROD-CD(W.SUB)
SET WT.CUST-CD = W.CUST-CD(W.SUB)
SET WT.EFCT-DT = W.EFCT-DT(W.SUB)
SET WT.TERM-DT = W.TERM-DT(W.SUB)
SET W.PROD-CD(W.SUB) = W.PROD-CD(W.SUB - 1)
SET W.CUST-CD(W.SUB) = W.CUST-CD(W.SUB - 1)
SET W.EFCT-DT(W.SUB) = W.EFCT-DT(W.SUB - 1)
SET W.TERM-DT(W.SUB) = W.TERM-DT(W.SUB - 1)
SET W.PROD-CD(W.SUB - 1) = WT.PROD-CD
SET W.CUST-CD(W.SUB - 1) = WT.CUST-CD
SET W.EFCT-DT(W.SUB - 1) = WT.EFCT-DT
SET W.TERM-DT(W.SUB - 1) = WT.EFCT-DT
ENDIF
ENDLOOP
Before
LOOP VARYING W.SUB
FROM W.ENTRIES BY -1 DOWN THRU 1
IF W.CURR-KEY < W.PREV-KEY
SET W.SUB2 = W.SUB - 1
SET WT.ARRAY-ROW =
W.ARRAY-ROW(W.SUB) BY POSITION
SET W.ARRAY-ROW(W.SUB) =
W.ARRAY-ROW(W.SUB2) BY POSITION
SET W.ARRAY-ROW(W.SUB2) =
WT.ARRAY-ROW BY POSITION
ENDIF
ENDLOOP
After

Ideal programming gotchas - $TRIM
§ $TRIM is one of the most CPU-intensive functions of CA Ideal
– When building output text strings
§ Avoid repeatedly trimming trailing blanks and so on from the same input
strings
– If repeatedly referencing fixed-length input
§ Convert to variable-length working data field with $TRIM just once and
reference that

Ideal programming gotchas - $TRIM (cont’d)
§ Appending data to an output “message” string
– Define output message string as variable-length working data field; no
need for $TRIM
– Use the CA Ideal 15.0 DOCUMENT feature
§ Particularly relevant to these types of applications
– Interfacing with IBM MQ-Series
– Ideal/web interface output documents

Re-Initializing Arrays
§ In batch, storage is cheap, CPU is expensive
– If an array has to be cleared many times during execution of a program
§ Keep an empty copy of the array in working data
§ Use SET real-array = empty-array BY POSITION
§ In CICS, storage is not so cheap, so this method may not be
appropriate for large arrays

Calling Non-Ideal Subprograms
§ If using any Language Environment (LE)-enabled language
– Make sure you are using the CEEUOPT RTEREUS=ON option
§ Ensures that LE runtime environment is only established once
§ Working data and so on is preserved between calls of the subprogram
§ Significant performance improvements where subprogram is called
repeatedly

Insert or update existing row
Determine which to use based on likelihood that row already exists
FOR FIRST dataviewname
WHERE...
update values
WHEN NONE
FOR NEW dataviewname
insert values
WHEN DUPLICATE
handle “unique” failures
ENDFOR
ENDFOR
FOR NEW dataviewname
insert values
WHEN DUPLICATE
FOR FIRST dataviewname
WHERE...
update values
ENDFOR
ENDFOR
Row is usually present Row is usually NOT present

Dataviews and elements
§ Dataviews are a collection of elements
– Elements are the unit of transfer
§ Defined as a single field
– costs vs benefits – now on zIIP
– Isolation from other table changes – Agile/DevOps
– SQL effectively does this
§ Defined as the whole row
– Use where the whole row is/will be ALWAYS required
§ Defined as all the columns in a key
– Index-only processing.
Elements are a Datacom-only concept

Available Tuning tools and resources

Built-in tuning tools
§ PXX Job statistics
– Shows number of Database requests and I/O’s
§ RXX reports (and READRXX)
– Useful to track maintenance activity and times
§ Datacom Accounting
– Can be very low overhead if defined correctly
– Can be extremely detailed for use when required
§ Dynamic System Tables
Included with the Datacom environment

Built-in tuning tools (cont’d)
§ AUTOINFO/AUTOSTATUS/AUTOCOLLECT
– Useful in spotting trends and contentions
§ MUF
– EOJ SYSPRINT
– JES LOG
§ CBS Diagnostic trace
§ SQL plan message options (or traces)

Built-in tuning tools (cont’d)
§ CICS DEBUG Facility
§ CA Ideal DEBUG facility
– Yes it runs in batch, and in runtime-only environments
§ IBM Debug Tool
– Basic but can be very useful

CA Application Tuning tools
§ CA Intertest
– Functionally rich
– Code Counting feature to identify heavily used statements
§ CA Sysview
– Excellent for analysis of current activity
– CICS Transaction Log provides historical analysis
§ Make sure Sysview’s DCCTXPR has been implemented
Available for use with the Datacom environment

CA Application Tuning tools (cont’d)
§ CA Mainframe Application Tuner (MAT)
– Major new tool for CA Ideal and COBOL and others
– Virtually no overhead and no recompiles required
§ I recommend you attend the various sessions on these
products while you are at CA World
§ Come see us at the Datacom demo station
Available for use with the Datacom environment

External resources
§ CADRE on CA Communities
communities.ca.com/community/ca-datacom-cadre
– There is no such thing as a “stupid question”
– Ask questions and share your experiences with developers across the
globe
– Submit your “Ideas” for product enhancements
– Specific enhancement requests are welcome
– Non-specific “I wish somehow..” requests are just as welcome
Available to CA Datacom users

External resources (cont’d)
§ CA Support
– support.ca.com/irj/portal/anonymous/newhome
– The friendly CA support staff have decades of experience and are
always happy to help
Available to CA Datacom users

Our results when applying application tuning
We have more Datacom tables than ever before.
We have more rows in those tables than ever before.
We have more applications processing the data than ever before.
We have faster response times (batch, Server and CICS) than ever before.
We have 30% less CPU (MSUs) than at the start of the tuning exercise.
The mainframe is no longer the highest TCO server.
Summary
Comments from the managing DBA for the two sites profiled in Part I

@CAWORLD #CAWORLD © 2016 CA. All RIGHTS RESERVED.55 @CAWORLD #CAWORLD © 2016 CA. All RIGHTS RESERVED.55
Mainframe and Workload Automation
For more information on Mainframe and Workload Automation,
please visit: http://cainc.to/9GQ2JI

Recommended Sessions
SESSION # TITLE DATE/TIME
MFX92S Strategy and Vision for CA Datacom and CA Ideal 11/16/2016 at 12:45 pm
MFX93S What’s new in CA Datacom 11/16/2016 at 1:45 pm
MFX94S Reducing TCO with Datacom Compression (Presspack) 11/16/2016 at 3:00 pm
MFX95SA Modernizing the CA Datacom data – SQL Access Part I 11/16/2016 at 3:45 pm
MFX95SB Modernizing the CA Datacom data – SQL Access Part II 11/16/2016 at 4:30 pm
MFX96S CA Datacom/AD Installation/Upgrade and Maintenance 11/17/2016 at 12:45 pm
MFX97S
CA Datacom/AD Report Utilization, and
information gather for Support
11/17/2016 at 1:45 pm
MFX98S
Using Sysview to monitor and measure your CA Datacom
environment – AD and DB
11/17/2016 at 3:00 pm
MFX99S Birds of a Feather/Stump the Techie! 11/17/2016 at 3:45 pm

Questions?

Thank you.
Stay connected at communities.ca.com

© 2016 CA. All rights reserved. All trademarks referenced herein belong to their respective companies.
The content provided in this CA World 2016 presentation is intended for informational purposes only and does not form any type of
warranty. The information provided by a CA partner and/or CA customer has not been reviewed for accuracy by CA.
For Informational Purposes Only
Terms of this Presentation

Application Tuning for CA Datacom Environments Part II

More Related Content

What's hot (20)

Viewers also liked (20)

More from CA Technologies (20)

Recently uploaded (20)

Application Tuning for CA Datacom Environments Part II