Application-enabling features of
DB2 10 and 11 for z/OS & DBA
enabling features
New England DB2 User’s Group
June 18, 2015
Jim Brogan
DB2 for z/OS Advisor
jambrog@us.ibm.com
© 2015 IBM Corporation
DB2 z/OS Availability Summary
2
Version
PID
General
Availability
Marketing
Withdrawal
End of
Service
5
5655-DB2
June 1997
December
2001
December
2002
6
5645-DB2
June 1999
June 2002
June 2005
7
5675-DB2
March 2001
March 2007
June 2008
8
5625-DB2
March 2004
September
2009
April 2012
9
5635-DB2
March 2007
December
2012
June 2014
10
5605-DB2
October 2010
July 2015
September
2017
11
5615-DB2
October 2013
http://www.ibm.com/software/data/support/lifecycle/
The aim of this presentation
 To help ensure that you are aware of recently
delivered DB2 for z/OS features that can boost
agility and productivity with respect to application
development
2
© 2015 IBM Corporation
Agenda (1)
 DB2 10 application-enabling features
– Temporal data support
– Enhanced SQL user-defined functions
– RETURN TO CLIENT cursors
– OLAP moving aggregates
– LOB enhancements
– Implicit casting of character string and numeric values
– Timestamp extensions
– XML enhancements
3
© 2015 IBM Corporation
Agenda (2)
 DB2 11 application-enabling features
– Autonomous native SQL procedures
– Array parameters (and variables) for SQL procedures
(and user-defined functions)
– Temporal special registers and temporal support for
views
– Global variables
– Transparent DB2-managed data archiving
– New grouping options: GROUPING SETS, ROLLUP,
CUBE
– DB2 integration with Hadoop-managed data
4
– XQuery support for XML data
© 2015 IBM Corporation
Application enabling features
of DB2 10 for z/OS
5
© 2015 IBM Corporation
Temporal data support
 Allows you to give a time dimension to data in a DB2 table
 Two flavors:
– System time: DB2 maintains a history table associated with a base table,
and will insert into the history table the “before” version of a row every
time a base table row is changed via update or delete
• DB2 also maintains “from” and “to” timestamp values in base and history table
rows, showing when a row in the history table was current, and when a row in
the base table became current
– Business time: a dimension that shows when data in a row is valid from
a business perspective (e.g., a product price that will go into effect next
year)
• You maintain business time values, but DB2 can help by preventing FROM
and TO business time period “overlaps” (so one version of a given row will be
valid from a business perspective at any given time)
– You can combine system and business time in one table (“bi-temporal”)
6
© 2015 IBM Corporation
More on temporal data support
 SELECT syntax extended to include the time dimension of a table
 Example: “What was the coverage associated with insurance
policy number 127348 at 10 AM on February 24, 2010?”
Can specify BUSINESS_TIME if table has
that dimension
Alternatively, can specify FROM and TO, or
BETWEEN two timestamp values
SELECT COL1, COL2,,,
FROM POLICY
FOR SYSTEM_TIME AS OF TIMESTAMP ‘2010-02-24
10.00.00’
WHERE POLICY_NUM = ‘127348’;
7
© 2015 IBM Corporation
Advantages of temporal data support
 System time makes it easy to provide an audit
history of data changes in a DB2 table
 Business time enables “forward looking” data
analysis possibilities
– Real-world example: forecasting future profit margins using
prices that will go into effect at a later time
 DB2-provided temporal capabilities GREATLY
increase programmer productivity versus “do it
yourself” temporal data functionality
 DB2-implemented temporal table functionality
delivers better performance than the do-it-yourself
alternative
8
© 2015 IBM Corporation
Enhanced SQL user-defined functions (UDFs)
 Prior to DB2 10, the “logic” in a SQL scalar UDF was restricted to
what you could code in the RETURN part of CREATE FUNCTION,
and that was quite limited
– RETURN could not contain a SELECT statement
– RETURN could not include a column name
 You were basically limited to receiving a value (or values) as
input, transforming that value (or values) arithmetically and/or
with scalar functions, and returning the result of that
transformation
– Example:
CREATE FUNCTION KM_MILES(X DECIMAL(7,2))
RETURNS DECIMAL(7,2)
LANGUAGE SQL
…
You can still create a UDF like this one,
but DB2 10 enabled you to do much more
with UDFs written in SQL
RETURN X*0.62;
9
© 2015 IBM Corporation
Enhanced SQL UDFs (continued)
 Starting with DB2 10, the RETURN part of a SQL scalar UDF can
contain a scalar fullselect
RETURN(SELECT WORKDEPT FROM EMP WHERE EMPNO = P1);
 Also new with DB2 10: the RETURNS part of a SQL scalar UDF can
contain a compound SQL statement, in which variables can be
declared and which can include logic flow control statements such
as IF and WHILE
BEGIN
DECLARE VAR1, VAR2 CHAR(10);
SET VAR1 = …;
IF P1 = …;
RETURN VAR2;
END@
 Also new with DB2 10: SQL table UDFs, which return a result set
10
© 2015 IBM Corporation
RETURN TO CLIENT cursors
 Prior to DB2 10, a cursor in a stored procedure could be declared
WITH RETURN TO CALLER, allowing the result set rows to be directly
fetched only by the direct caller of the stored procedure
– Example: program PROG_A calls stored procedure PROC_B, which
calls procedure PROC_C, which has a WITH RETURN TO CALLER
cursor
– PROC_B can directly fetch rows from the cursor, but PROG_A cannot
• If PROG_A needs the result set, PROC_C can put it in a temporary table, and
PROG_A can get the rows from that temp table
– Clunky from a programming perspective, and not optimal for performance
 DB2 10: stored procedure can declare a cursor WITH RETURN TO
CLIENT
– Makes result set rows directly FETCH-able by “top-level” program (i.e.,
the one that initiated a chain of nested stored procedure calls)
11
© 2015 IBM Corporation
Previous slide’s point, in a picture…
 Before DB2 10:
Program XYZ
WITH RETURN TO CALLER
CALL
Stored proc A
CALL
Stored proc B
DECLARE C1 CURSOR…
 With DB2 10:
WITH RETURN TO CALLER
Program XYZ
Still an option with
DB2 10, if this is the
behavior you want
CALL
Stored proc A
CALL
Stored proc B
WITH RETURN TO CLIENT
12
DECLARE C1 CURSOR…
© 2015 IBM Corporation
OLAP moving aggregates
 A new (with DB2 10) SQL syntax that allows:
– Partitioning of a result set (e.g., by name)
– Ordering of rows within result set partitions (e.g., by date)
– Generation of aggregate values based on the “moving” current
position within a set of rows (e.g., sum of sales for the current
row plus the two preceding rows)
The desired result set
– Example:
partitioning value
SELECT NAME, DATE, UNITS_SOLD,
SUM(UNITS_SOLD) OVER(PARTITION BY NAME
ORDER BY DATE
function
ROWS BETWEEN 2 PRECEDING AND CURRENT ROW) SUM
FROM PRODUCT_SALES;
13
rows within result set
partitions
The desired scope of aggregation as DB2
moves through the result set partitions
© 2015 IBM Corporation
The result of the SELECT on the previous slide
14
NAME
DATE
UNITS_SOLD
SUM
Jones
2015-01-10
7
7
Jones
2015-01-11
8
15
Jones
2015-01-12
5
20
Jones
2015-01-13
6
26
Smith
2015-01-10
4
4
Smith
2015-01-11
9
13
Smith
2015-01-12
8
21
Smith
2015-01-13
5
26
Sum of this row’s
UNITS_SOLD (5) plus the
UNITS_SOLD values of the
preceding two rows in the
result set partition (7 and 8)
© 2015 IBM Corporation
LOB enhancements: Inline Lobs
 Prior to DB2 10, every bit of every value in a LOB column had to be
physically stored in a separate LOB table space (the LOB values
logically appear to be in the base table rows)
 With DB2 10, a LOB column’s definition can include a specification
of the amount of space in the base table that can be occupied by
LOB values
– The portion (if any) of a value over the limit is stored in LOB table
space
 Great for a LOB column for which relatively few values are truly large
– Can significantly improve the performance of LOB-reading and LOBinserting programs (and utilities) when most of a LOB column’s values
can be completely in-lined
– Also allows creation of index on expression on in-lined portion of a
CLOB column (using the SUBSTR)
15
• Example: if contracts are stored in a CLOB column, and if data in bytes 10
through 20 is always the contract number, can build index on that
© 2015 IBM Corporation
Lob enhancements: utilities
 Variable-block spanned (VBS) record format now supported for data
sets used for table UNLOAD and LOAD (referring to SYSREC data
set)
– What this means: you can unload a table with a LOB column (or
columns) and have ALL of the data – LOB and non-LOB – go into a
single data set
– And reverse is true for LOAD (i.e., data – LOB and non-LOB values –
can be loaded from a single input data set
• Before DB2 10, had to unload individual LOB values to members of a PDS,
or to individual files in the z/OS UNIX System Services file system (and
reverse was true for LOAD)
• DB2 10 spanned record support greatly simplifies use of UNLOAD and
LOAD for tables with LOB columns, and substantially boosts performance
 DB2 10 also delivered support for online REORG of LOB table space
with SHRLEVEL(CHANGE)
16
© 2015 IBM Corporation
Implicit casting of character, numeric values
 Consider this statement:
SELECT 1 CONCAT ‘+’ CONCAT 1 CONCAT ‘=‘ CONCAT 2
FROM SYSIBM.SYSDUMMY1;
 In a pre-DB2 10 environment, that statement gets this result:
SQLCODE = -171, ERROR: THE DATA TYPE, LENGTH,
OR VALUE OF ARGUMENT 1 OF CONCAT IS INVALID
 In a DB2 10 (new-function mode) system, you get this:
•1+1=2
 Works assignment (SET) statements, too (but not for special
registers)
 Numeric values are implicitly cast to VARCHAR, character values
are implicitly cast to DECFLOAT(34)
– Why? Because VARCHAR and DECFLOAT(34) are compatible
with all other character and numeric data types, respectively
17
© 2015 IBM Corporation
Timestamp extensions
 New with DB2 10: timestamp values down to the picosecond
(that’s a trillionth of a second)
– One reason this was needed: mainframe engines are so fast now
that microsecond-level timestamps (often defined as unique keys
in DB2 tables) can regularly produce duplicate values
 Also new with DB2 10: variable-precision timestamps
– From 0 (no fractions of a second) to 12 (picosecond-level
precision), with 6 being the default
– Syntax: TIMESTAMP(n)
 Another DB2 10 enhancement: TIMESTAMP(n) WITH TIME ZONE
– New data type
– Sample value: ‘2012-10-03-10.15.00.123456-05:00’
18
Difference between
local time and UTC
© 2015 IBM Corporation
XML enhancements
 With DB2 10, you can specify in the definition of a table the XML schema
that is to be used to validate data inserted into an XML column
– No longer have to invoke DB2-supplied user-defined function to accomplish
schema validation
– Additionally, DB2 10 XML schema validation is done “in the DB2 engine”
• Better performance, and zIIP-eligible
 And, you can update part of an XML document (versus replacing the whole
thing) via new XMLMODIFY built-in function
– Can insert a node into an XML document, replace a node, delete a node, or
replace values of a node
 Also, the CHECK DATA utility can check on the structural validity of XML
documents in an XML table space
– Pre-DB2 10: only checked consistency between base table and XML table
space
19
© 2015 IBM Corporation
Application enabling features
of DB2 11 for z/OS
20
© 2015 IBM Corporation
Autonomous native SQL procedures
 A DB2 11 native SQL procedure can function as an autonomous
transaction
– How it’s done: AUTONOMOUS option specified in CREATE
PROCEDURE (or ALTER PROCEDURE) statement
• Specified instead of COMMIT ON RETURN YES/NO
– An autonomous SQL procedure commits on returning to the calling
program, but (unlike the case when COMMIT ON RETURN YES is in
effect) that commit does NOT affect the calling program’s unit of work
– An autonomous SQL procedure’s unit of work (UOW) is independent of
the calling program’s UOW – if the calling program’s UOW is rolled
back, data changes made by autonomous SQL procedure will not be
rolled back
• Very useful if you require that a data update be accomplished when a
transaction executes, and you need that update to persist even if the
transaction subsequently fails
– A restriction: one autonomous SQL procedure can’t call another
21
© 2015 IBM Corporation
Array parameters (and variables) for SQL
procedures (and UDFs)
 DB2 11: array parameters can be passed to (and/or received
from), and array variables can be declared in, native SQL
procedures (and the same is true for SQL user-defined
functions)
– Call to SQL procedure with array input or output parameter can
come from a SQL PL routine, a Java program, or a .NET
program (for latter two, via IBM Data Server Driver type 4 driver)
• If .NET caller, array must be input parameter
– An array in this context is a form of a DB2 user-defined data type
(UDT) – you create it, then you use it
– Built-in functions are provided to:
•
•
•
•
22
Construct arrays
Derive tables from arrays
Obtain information about arrays
Navigate array elements
© 2015 IBM Corporation
More on array parameters and variables
 There are two array types:
– Ordinary
• Has a user-defined upper bound on number of elements (defaults to
INTEGER high value)
• Elements referenced by their ordinal position in the array
– Associative
• No user-defined upper bound on number of elements
• Elements are ordered by and can be referenced via array index values
• Values in a given array index are INTEGER or VARCHAR, are unique, and
don’t have to be contiguous
This is an ordinary array – associative array would have data type of index
values (e.g., VARCHAR(8)) after ARRAY keyword
CREATE TYPE PHONENUMBERS AS DECIMAL(10,0) ARRAY[50];
Data type of values in the array
23
Max number of elements in ordinary
array (defaults to about 2 billion)
© 2015 IBM Corporation
Temporal special registers
 The need: what if you want a program (or just a SQL statement) to have
an other-than-current view of temporal data, but you don’t want to
change the program’s code?
 Solution: two new special registers delivered with DB2 11
– CURRENT TEMPORAL SYSTEM_TIME
– CURRENT TEMPORAL BUSINESS_TIME
 When set to a non-null value, has the effect of adding the following to a
SELECT statement that targets a temporal-enabled table (in this case, use
of system time is assumed):
– FOR SYSTEM_TIME AS OF CURRENT TEMPORAL SYSTEM_TIME
 Example of setting special register’s value:
SET CURRENT TEMPORAL SYSTEM_TIME = CURRENT TIMESTAMP – 1 YEAR;
(this would result in a program having a view of data that was current as of one year ago)
24
© 2015 IBM Corporation
More on temporal special registers
 A special register non-null value, once set, remains in effect for
that particular session (thread) until it’s changed (setting to null
has the effect of “turning the special register off”)
– But if set within a routine (stored procedure or UDF), the new value is not
passed back to the invoking application
 SYSTIMESENSITIVE, BUSTIMESENSITIVE bind options determine
whether or not SQL statements (static or dynamic) issued
through a package will be affected by temporal special registers
– Default value is YES
 If CURRENT TEMPORAL SYSTEM_TIME is set to non-null value
for a thread, data modification statements targeting system timeenabled tables are not allowed
25
© 2015 IBM Corporation
Temporal support for views
With DB2 11, you can use temporal predicates when referring to a view defined
on a temporal table (but you can’t use a temporal predicate in defining a view)
SQLCODE -4736
Base table
Temporal predicate
View
Temporal predicate
26
© 2015 IBM Corporation
Global variables
 The need: how can you pass data values from one SQL statement to
another in the context of a thread?
– Before DB2 11:
• Do it with application code (values placed into variables by one SQL
statement are copied to variables used as input to another SQL statement)
• Want a trigger to be able to access those values? Not easy…
– DB2 11: use global variables
 You can create your own global variables using the new CREATE
VARIABLE statement
– DB2 11 also provides a few built-in global variables:
• SYSIBM.CLIENT_IPADDR
• SYSIBMADM.GET_ARCHIVE
• SYSIBMADM.MOVE_TO_ARCHIVE
27
More on this archive stuff
momentarily…
© 2015 IBM Corporation
Global
variables
example
Reference the global variable
Assign value to a
(previously created)
global variable
28
© 2015 IBM Corporation
Transparent DB2-managed data archiving
Meaning: infrequently
referenced
 The need: get old and “cold” data out of a table (for better SQL and
utility performance), but retain deleted rows in an archive table and
transparently enable retrieval of archived rows
 DB2 11 will do this for you
 What a DBA does (suppose that table TAB_A is to be archiveenabled):
– Create an archive table that looks just like TAB_A (same number of
columns, same order of columns, same column names and
definitions)
– Tell DB2 to associate the archive table with TAB_A (assume that you
named the archive table TAB_A_ARCHIVE):
ALTER TABLE TAB_A ENABLE ARCHIVE USE TAB_A_ARCHIVE;
29
© 2015 IBM Corporation
More on DB2-managed data archiving
 Temporal and archive tables are mutually exclusive
 New built-in global variables affect interaction with archive-enabled
tables (default value for both is ‘N’):
– SYSIBMADM.GET_ARCHIVE – if ‘Y’ then SELECT targeting
archive-enabled table will automatically include UNION ALL with
archive table
– SYSIBMADM.MOVE_TO_ARCHIVE – if ‘Y’ or ‘E’ then rows deleted
from archive-enabled table will be inserted into archive table (if ‘Y’
then INSERT/UPDATE/MERGE disabled for base table)
 ARCHIVESENSITIVE bind option determines whether statements
(static or dynamic) will be affected by value of
SYSIBMADM.GET_ARCHIVE global variable (default is YES)
– ARCHIVE SENSITIVE option on create of native SQL procedure or
UDF does the same thing
30
© 2015 IBM Corporation
New grouping option: GROUPING SETS
Example: determine average total compensation for WorkDept, Job, and EdLevel sets
Basically means, “group
by each of these
columns, in turn”
31
© 2015 IBM Corporation
New grouping option: ROLLUP
Example: determine average total compensation for the various hierarchies of
WorkDept, Job, and EdLevel, and for overall set
• Column order in GROUP BY expression affects result set
• ORDER BY helps with readability
You get a grouping by all values of
column 1, column 2, and column 3; a
grouping by all values of column 1
and column 2; and a grouping by all
values of column 1
32
You also get an aggregate
over all qualifying rows
© 2015 IBM Corporation
New grouping option: CUBE
 Example: determine average total compensation for various combinations of
WorkDept, Job, and EdLevel
– Column order in GROUP BY expression doesn’t matter
– ORDER BY helps with readability
You get grouping by all values of all three
columns, by all values of all combinations
of two of the three columns, and by all
values of each individual column
You also get an aggregate over all qualifying rows
33
© 2015 IBM Corporation
DB2 integration with Hadoop-managed data
 Hadoop: an open source software framework that supports data-intensive
distributed applications
 Two main components
–Hadoop distributed file system
–MapReduce engine
• Powerful, but tedious from a development
perspective
• “Like the assembly language of Hadoop”
34
© 2015 IBM Corporation
DB2 11: new UDFs for Hadoop integration
Available now for
Linux on z Systems
A new user-defined function (UDF)
allows a data analytics job,
specified in JAQL, to be submitted
to a BigInsights server
IBM BigInsights
A new table UDF reads the output
of the analytics job and returns it
in relational form
35
© 2015 IBM Corporation
XQuery support for XML data
 Pre-DB2 11: XPath expressions can be used to navigate
through XML documents and to address parts of XML
documents
– XPath is a subset of XQuery, which is a richer language for
accessing XML documents
– XPath limitations often necessitated using a mixture of XPath
and SQL, and that could make query coding more difficult
 DB2 11 includes XQuery support, providing a richer set of
XML expressions that can be used with the built-in functions
XMLQUERY, XMLEXISTS, and XMLTABLE
– Queries can be expressed purely using XQuery, versus a
mixture of XPath and SQL, and that can boost programmer
productivity
36
 XQuery support was retrofitted to DB2 10 via APARS
PM47617 and PM47618
© 2015 IBM Corporation
In conclusion…
 DB2 10 and 11 delivered a lot of new applicationenabling features
– How many of these are being used at your site?
– How many could be put to good use at your site?
37
© 2015 IBM Corporation
Migration Catalog / Directory Summary
•
All new non-AUX create Table Spaces are PBGs
– MAXPARTITIONS 1
– DSSIZE 64G
•
Catalog / Directory Visualization
39
108 Unicode; 7 EBCDIC; 1 ASCII
 TS Page Size Counts
–
–
–
–
4K = 76 -> 94
8K = 10
16K = 5 -> 6
32K = 4 -> 6
Application Compatibility ...
• Allows applications to continue experiencing SQL
DML & XML behavior from a previous release
(DB2 10)
– APPLCOMPAT ZParm
• Must be V10R1 until NFM
– Migrations default to the migrate-from release
(V10R1)
– Installations default to the migrate-to release (V11R1)
• DDL and DCL is not fenced by APPLCOMPAT
• This includes new functions, not just changes to
existing behavior
• May not be possible when conforming to SQL
standards
40
Application Compatibility ...
• Similar limited capability in DB2 10
– BIF_COMPATIBILITY
– DDF_COMPATIBILITY (Disabled when APPLCOMPAT set to (V11R1) via PM94719
• Static SQL is governed by the Package APPLCOMPAT
– Cannot bind with V11R1 until NFM
• Dynamic SQL is governed by the:
– CURRENT APPLICATION COMPATIBILITY
– Which defaults to the Package APPLCOMPAT if not SET
• Once in NFM:
– CURRENT APPLICATION COMPATIBILITY can be set to either level
– RE/BIND PACKAGE can choose either level
41
Application Compatibility ...
• BIND / REBIND (TRIGGER) ... APPLCOMPAT Defaults
– Must be V10R1 until NFM
• V11R1 before NFM returns error
– During REBIND, if already rebound with a compatibility level
• Defaults to this level
• Including Autobind
– If BINDind, or not yet set, defaults to APPLCOMPAT
ZParm
• CREATE / ALTER PROCEDURE /FUNCTION
– Same rules as BIND / REBIND
• On DB2I defaults panels for
– BIND: DSNEBP10
– REBIND: DSNEBP11
42
Application Compatibility ...
• CURRENT APPLICATION COMPATIBILITY
– Defaults to the Package RE/BIND
– If not rebound, defaults to the ZParm
– SET this Special Register overrides all
• Must be in NFM to SET this register
• IFCID 239
– Indicates Packages using a function that changes in DB2 11
– Field QPACINCOMPAT
– See SDSNMACS(DSNDQPAC) for mapping
• IFCID 366/376
– Records indicate SQL using the V10 code path which is different from the V11 code path
– Use these in CM to identify programs needing review
– 376 is new in V11 and is a roll up of activity reported in 366
• Attempts once per dynamic and static statement (bound V10 or later)
• Once per Plan, Package, Statement # bound prior to V10
– See SDSNMACS(DSNDQW05) for detailed description
43
Application Compatibility ...
10 NFM
•
•
•
11 ENFM
11 CM
ZParm V10R1
 Same as CM
– Can set to
V11R1 but will
not operate
that way
BIND/REBIND
– Must be V10R1
CREATE/ALTER
– Must be V10R1
•
•
SET CAC* not
available
IFCID 239/366/376
11 NFM
 ZParm
– V10R1 or V11R1
 BIND/REBIND
– V10R1/V11R1 available
 BIND
– Defaults to ZParm
 REBIND & Autobind
– Defaults to previous
Catalog value first
– ZParm second
 CREATE/ALTER
– V10R1 or V11R1
 SET CAC* available
 New features
– Require V11R1
*CAC = CURRENT APPLICATION COMPATIBILITY
44
RBA / LRSN Expansion ...
• DB2 10 and prior used a 6 byte (Basic) RBA / LRSN
– Some customers have had to take RBA action to keep
systems running
• RBA reset
– Manual recovery action to reset
• Bring up new data sharing members
– Represents 256 TBs of logging space
– Messaging (DSNJ032I) and system actions
• F000 0000 0000 warning threshold surfaces at log switch & restart
• FFFF 0000 0000 critical threshold, DB2 will only run
ACCESS(MAINT)
– The LRSN also has a limit in Year 2042 if there’s no DELTA
RBA
– STORCLK is 8 bytes, and therefore LRSN spin can occur
45
RBA / LRSN Expansion ...
• DB2 11 offers the option to convert to a 10 byte RBA / LRSN
– Almost 4 Billion times the logging space
– RBA extended with high-end (left) 4 bytes
• Extended value represents 1 YB of logging space (1 YB = 1 Trillion TBs)
– LRSN
• 1 byte on high-end
• 3 bytes on the low-end
• Adding 30,000 years of logging
– Conversion is optional if not nearing the 6 byte limit
• However, DB2 11 uses 10 bytes internally and converts to 6 on writes
– Convert earlier than required to:
• Avoid internal conversions
• Resolve LRSN spin conditions
• Disabling Data Sharing requires 10 byte RBA/LRSN for surviving members
46
RBA / LRSN Expansion ...
•
•
Must restart the subsystem / member on NFM before expansion
There are 3 areas for conversion which can be done in any order and schedule IF
NOT NEARING 6 byte limit
– BSDS (DSNTIJCB)
• DB2 is down
• BSDSs are reformatted and will grow
• A checkpoint will be taken at restart
– Catalog (DSNTIJCV)
• Some of these may have been converted during DSNTIJEN (ENFM)
– User Data / Indexes
• REORG, REBUILD, LOAD REPLACE with any SHRLEVEL
• At the PART level, except for XML* & Hashed
• Data & Indexes can be converted independently
• Growth is not expected, as there is enough room in the pages
•
DB2 11 internals are always using a 10 byte RBA / LRSN
– But can be converted to 6 byte for objects in BASIC format for writes
– As long as DB2 is not close to the 6 byte limit
* Research options for XML. Based on the options and if previously was Expanded, PART may be an option
47
• DSNTIJCB
RBA / LRSN Expansion ...
– Tailored during the MIGRATE execution of the CLIST
– DB2 must be down
• Recommend –STOP DB2 MODE(QUIESCE)
– Defines new BSDSs (suffix NEWFMT1)
– Converts into the new BSDS data sets
– Renames:
• existing to .OLDFMT
• NEWFMT1 to existing
– A demo system observed a 50% growth in the BSDS
48
• DSNTIJCV
RBA / LRSN Expansion ...
– Tailored during the ENFM execution of the CLIST
– Consider MODIFY RECOVER & REORG prior to conversion
• For large / active Catalogs
• When executed outside of the migration project to clean up again
– All converted areas are COPYed
– As with JEN, SYSUTILX cannot be REORGed and therefore is reset
• Make sure it is empty and quiesced before running JCV
• CATMAINT converts SYSUTILX
– REORG ... SHRLEVEL CHANGE converts the rest
– Builds LISTDEF with BASIC YES, so already converted objects are skipped
– This job can be split apart, but review PM95430
• When SYSLGRNX is Extended & SYSCOPY is Basic
• COPY SHRLEVEL REFERENCE is not recorded
49
• In CM:
RBA / LRSN Expansion ...
– IFCID 306 & Log Capture Exit products (Replication)
see 10 bytes
– Work files are Extended when opened
– DISPLAY THREAD and restart status messages are 10
bytes
• In ENFM
– JEN REORGed Catalog / Directory objects converted
based on UTILITY_OBJECT_CONVERSION
• In NFM
– JNF converts SCA to 10 bytes
– DSNTIJCB & DSNTIJCV available to convert BSDS &
Cat/Dir
– Use utilities RBALRSN_CONVERSION keyword for user
data/indexes
50
RBA / LRSN Expansion ...
• Clone tables must be dropped to convert
• SYSTABLEPART & SYSINDEXPART indicate format (RBA_FORMAT)
– All migrated objects will indicate blank
– At open time, REORG, REBUILD, LOAD, or REPAIR these values will be set as “B” or “E”
– CREATE sets RBA_FORMAT
• DEFINE NO = “U”
– DEFINE NO objects created according to OBJECT_CREATE_FORMAT in effect
• Utilities and Service Aids accept either format
– UTILITY_OBJECT_CONVERSION defaults utility RBALRSN_CONVERSION keyword.
– DSN1COPY may need REPAIR to correct the Catalog definition
• Must be TO a DB2 11 NFM system if Expanded
– RECOVER of Expanded object with a Basic copy will reset the object to Basic
• Cannot roll forward across an RBA / LRSN change
•
IFCIDs
– Many are reformatted.
– The RBA/LRSN was moved when possible to not impact other fields
51
•
RBA / LRSN Expansion
V11 6 byte Thresholds
– RBA
• F000 0000 0000
– DSNJ032I Warning with each log switch and DB2 restart
• FFF8 0000 0000
– -904 00C2026D Soft Limit
– No SQL Updates allowed for Basic Objects
• FFFF 0000 0000
– DSNJ033E Critical Threshold
– DB2 stops with 00D10251
– DB2 only runs in ACCESS(MAINT)
• If you manage to reach all ‘F’s, then previous RBA reset procedure is still likely
– LRSN
• 1 Year left
– DSNJ034I Warning with each log switch and DB2 restart
• 2 Months left
– 904 00C2026D Soft Limit
– No SQL Updates allowed for Basic Objects
• 2 Weeks left
– DSNJ033E Critical Threshold
– DB2 stops with 00D10251
– DB2 only runs in ACCESS(MAINT)
52
Suppress-null Indexes
 Support creation of indexes where the default NULL values are excluded from the
index
– Reduced index size, improved insert/update/delete performance, 3rd party DBMS
compatibility
 New CREATE INDEX keywords
– INCLUDE NULL KEYS (default)
• DB2 will create an index entry even if every key column contains the NULL value
– EXCLUDE NULL KEYS
• DB2 will not create an index entry when every key column contains the NULL value. If any key column
is not null the index entry will be indexed
 Conditions:
– EXCLUDE NULL KEYS must not be specified if any of the columns identified by column-name
are defined as NOT NULL
– Not supported if the index is defined as a partitioning index
53
IDX
Auto Cleanup of Pseudo-deleted Index Entries
 DB2 11 automatically cleans up pseudo deleted entries
– zIIP eligible and runs in the background
– Designed to have minimal or no disruption to applications
– New ZParm INDEX_CLEANUP_THREADS (0-128) to control number of concurrent
cleanup tasks
–Default is 10
–0 disables index cleanup
–Value can vary between members of a data sharing group
–Online changeable
– New SYSIBM.SYSINDEXCLEANUP Catalog table to control auto cleanup at index
level
• Day of week/month, start/end time.
• By default cleanup is enabled for all indexes
– IFCID 377 written when index is cleaned up
 -DIS THREAD TYPE(SYSTEM) to see threads
– Correlation id in the output shows: 014.IDAEMKxx
– xx indicates thread number (01, 02…)
54
IDX
Pseudo-deleted Index Cleanup
 Cleanup done under system tasks, run as enclave SRBs and zIIP eligible
• Parent thread (one per DB2 member) reads through RTS to find candidates
–Runs every 5 minutes
• Eligible indexes sorted based on number of pseudo deleted rids to delete highest first
• Child threads assigned based on ZParm setting
SELECT FROM… ORDER BY
SYSIBM.SYSINDEXSPACESTATS
NAME
…
IX1
n
100 X
5000
IX2
n
1000 X
20000
IX3
n
500 X
100000
IX4
n
2000 X
75000
NPAGES
…
REORGPSEUDODELETES
– Child cleanup thread only started
if Index already open for INSERT,
UPDATE or DELETE
• ‘X’-type P-lock already held
55
IDX
Child cleanup
thread IX3
Child cleanup
thread IX4
Parent
thread
Index
IX3
IX4
IX2
IX1
In
Memory
SYSIBM.SYSINDEXCLEANUP Example
 All index spaces in DB_1234 are enabled for cleanup on Sundays from 4:30 until noon,
except
– Index space IX_9876 is always disabled for cleanup
SYSIBM.SYSINDEXCLEANUP
DBNAME
INDEXSPACE
ENABLE_
DISABLE
MONTH MONTH
_WEEK
DAY
START
_TIME
END
_TIME
DB_1234
NULL
E
W
NULL
7
043000
120000
DB_1234
IX_9876
D
NULL
NULL
NULL
NULL
NULL
 DB2 checks SYSIBM.SYSINDEXCLEANUP table at 10 min intervals
– Enforcement of new row may be delayed up to 10 min
 RECOMMENDATION: Use rows in SYSIBM.SYSINDEXCLEANUP only to define exceptions
to default index cleanup behavior
– Define time windows at system or database levels, rather than specific indexes when
possible
– Remove unneeded or conflicting rows
56
IDX
38
© 2015 IBM Corporation