Monday, October 11, 2021

What is Db2?

For regular readers of this blog the title of this particular post might seem to be a bit odd. After all, why visit the Db2Portal blog if you don't know what Db2 is, right?


Well, the purpose of this short post is to refer my regular readers to an article I wrote for TechTarget, a website that regularly posts "What Is" type of articles.

So, you guessed it, I wrote the latest, updated "What is Db2?" article for TechTarget... and I figured I'd share it with my blog readers. 

Now whenever somebody asks a high-level question about Db2, like "what is it?" or "what are the advantages and disadvantages of Db2" or even "where can I run Db2?" you now have a place to send them...

Here ---> What is Db2?

Monday, August 30, 2021

What Type of Changes Cause Db2 Packages to Get Invalidated?

Db2 DBAs are constantly working with database objects such as Databases, Tablespaces, Tables, and Indexes. And many requirements cause DBAs to have to modify these objects. Some modifications may be simple, such as just issuing an ALTER statement. Others may be more in-depth, even to the point of having to DROP and re-CREATE the object.

Now I've blogged here before about the many different types of Db2 database changes and how to manage them. For reference, here is a blog post that summarizes the posts I've made on this topic.

My purpose today is not to rehash all of that information again, but to discuss one aspect of change management that probably causes DBAs the most grief: package invalidation. 

Packages can be invalidated by Db2 for many reasons. For example, when a privilege that is required by a package is revoked, the package is invalidated by Db2. 

When a package is invalidated it cannot be executed until it has been rebound. This can be automatic, but it is usually better to be proactive and to Rebind packages when you take an action that invalidates packages. 

And as we all know, rebinding can cause access paths to change. Hopefully for the better... but not always. If access paths always got better then there would be no DBA grief, right? Whenever DBAs perform Rebinds they are always dreading that call from the developer or end-user that says "Hey, this transaction (or job) that used to run quickly is now taking forever."

So it makes sense that DBAs need to be aware of what changes cause packages to be invalidated. Of course, if you have to DROP an object that the package accessed it is obvious that a Rebind is required. But there are many other types of changes that will invalidate packages.

Fortunately, the IBM Db2 documentation is good and easy to find. Here is a link to the Db2 12 for z/OS documentation for Changes that invalidate packages. If you are a DBA, I recommend that you click on that link and bookmark that page!

I'm not going to copy and paste all of the information from the manual here (no reason to and over time it could change). But here are some of the things to keep in mind that you may not at first think will affect packages, but can:

  • Altering, dropping, or renaming a column
  • Adding date/time columns with defaults
  • Adding a constraint
  • Adding, changing, or rotating partitions in a partitioned or partition-by range UTS tablespace
  • Temporal and transparent archiving changes
  • Adding, altering, or dropping a materialized query table (MQT) 
  • Activating or deactivating row-level or column-level access control
  • Enabling or disabling masks if column access control is in effect
  • Increasing a table space's MAXPARTITIONS attribute
  • Changing a table space's SEGSIZE or DSSIZE
  • Changing the buffer pool for a tablespace (with a different page size)
  • Altering indexes to add a column, change the PADDED attribute, or changing the limit key value of a partitioning index
  • Regenerating an index
  • Running the REORG utility with the REBALANCE keyword
  • Running the REPAIR utility on a database with the DBD REBUILD option

Again, these are just some of the admin changes that can invalidate packages. There are others and you should always refer to the current Db2 documentation for the list of things that will invalidate packages before you make any changes. Failing to do so might mean that you will have to run a mass Rebind... maybe at a time when you'd rather not!

Finally, I'll leave you with a couple of helpful queries you can run to help as you manage changes.

To identify all packages that will be invalidated by a change to a specific object, run the following query:

SELECT   DISTINCT DCOLLID, DNAME, DTYPE 
FROM     SYSIBM.SYSPACKDEP
WHERE    BQUALIFIER = ?
AND      BNAME = ?
AND      BTYPE = ?
ORDER BY DCOLLID, DNAME;

Simply plug in the qualifier and name of the object, along with type of the object (appropriate values can be found in the Catalog table documentation in the appendix of the IBM Db2 SQL Reference manual).

And if you want to identify all invalid packages, try running this query:

SELECT   COLLID, NAME, VALID
FROM     SYSIBM.SYSPACKAGES
WHERE    VALID <> 'Y'
ORDER BY COLLID, NAME;

Monday, August 16, 2021

SQL to Return the nth Maximum Value

Sometimes what seems like a simple request might take a little bit of thinking to devise and implement a solution. Recently, I was asked how to write SQL that returns the nth maximum value from a column. If you think about it, it isn't too difficult.

Assume that you have a table with 10 rows, and there is a column in that table that holds the values 1 through 10. The idea is to create a query that for the 7th maximum value would return a 7, the 8th an 8, and so on.

This can be accomplished with a common table expression (CTE) and the following query:

WITH TOPN AS 
   (
    SELECT YOUR_COLUMN 
    FROM    YOUR_TABLE
    ORDER BY YOUR_COLUMN DESC
    FETCH FIRST 7 ROWS ONLY
   )
SELECT MIN(YOUR_COLUMN)
FROM   TOPN;

Simply change the value of FETCH FIRST from 7 to whatever you wish n to be.

If there are duplicate values and you want to eliminate them from consideration, just change the SELECT to SELECT DISTINCT in the TOPN common table expression.

This should work for any type of values (numeric, character, date/time) that you wish to query.

Thursday, July 29, 2021

New IBM Storage Systems Boost Ability to Gain Value from Your Mainframe Data

Gain more value from your mainframe data with IBM Storage

Every year the amount of data that is created continues to expand. Analysts at IDC estimate that data will grow at a compound annual growth rate of 23 percent through the year 2025Furthermore, efficient access to critical business data can mean the difference between success and failure, yet we sometimes forget about the crucial role that storage systems play in our everyday business transactions.

While storage systems have gotten more intelligent and fault-tolerant over the years, there’s always room for advances that can deliver an improved user experience. This can be seen by IBM’s latest storage announcements. The announcement highlights new and improved storage capabilities including cloud-like consumption models, data resiliency, and mainframe storage. This post will focus on the mainframe aspects of the announcement.

Why mainframe? Well, the platform continues to prosper and grow. According to the latest BMC Mainframe Survey, 90 percent of the IT leaders surveyed see the mainframe as a long-term platform for growth. The world’s largest organizations rely on the mainframe to deliver superior performance, reliability, and security. Mainframes are being used not just for traditional transaction processing and batch workloads, but also for new workloads running business analytics and AI applications on structured data. Not to mention that these large shops store most of their data on the mainframe!

What’s New

On July 20, 2021, IBM announced the next generation of its Storage for the IBM Z, the IBM DS8980F analytics class storage system. Engineered to excel for modern workloads that span transaction processing, analytical processing, and AI for native cloud and on-premises computing. The DS8980F offers high-speed and high availability as a single all-flash storage solution.

As part of this announcement, IBM is introducing improvements in Safeguarded Copy to the entire family of IBM DS8900F systems – including DS8910F, DS8950F and the new DS8980F – to greatly reduce the recovery time from a remote location to the production environment. Additionally, IBM is bringing the Safeguarded Copy function in IBM Spectrum Virtualize software to the IBM FlashSystem family and IBM SAN Volume Controller. 

Bringing the focus back to the mainframe: the new IBM DS8980F storage system has been developed by IBM with its z15 mainframe hardware in mind. That means it is optimized for mainframe-class workloads. Organizations are continually looking for ways to improve the performance of their mainframe applications, and the DS8980F provides the fastest mainframe application response times. Therefore, a key method of improving performance can be to upgrade your storage system. Indeed, the new IBM DS8980F, compared to the last generation of IBM storage systems (DS8888F series), can improve response time by up to 25 percent. 

Minimizing downtime is another critical requirement of modern business applications, especially for those that run on mainframes. The new IBM DS8980F delivers 7 nines of availability (99.99999 percent), an improvement of 10x over the previous generation.

Additional improvements include more than twice the amount of system cache and greater bandwidth capacity, all while requiring less energy consumption and in a lighter-weight box.

At the same time, IBM also announced a new tape library system, the IBM TS7770, with all flash cache. The most significant new feature of the TS7770 is that it provides better performance with only 1 flash drawer than the previous 10 SAS HDDs drawers, delivering faster data protection with less infrastructure.

Finally, it is possible to combine the TS7770 tape library and the IBM DS8910F  (the entry version within the DS8900F family) into a single 19-inch industry-standard rack. This enables smaller and medium-sized organizations to deploy an end to end storage solution for mainframe environments, into a smaller amount of floor space with important savings in operating costs.

Summary

Data growth continues unabated, and organizations continue to use mainframes expecting them to deliver unparalleled performance and availability for their mission-critical workloads of all types.

To achieve this level of performance and availability, while managing data growth, organizations need the latest and greatest storage technology. And IBM’s latest DS8980F and TS7770 will help organizations achieve the performance and availability they require for all their application workloads.

If you’d like to learn more about the latest from IBM storage, you can read the full details in the IBM announcement.


Friday, July 16, 2021

Keeping Track of Data Movement in Db2 for z/OS

Creating and managing test data for Db2 application development and testing requirements can be a significant challenge. To enable not only the development of new programs, but to be able to maintain existing ones, organizations must ensure that there is an adequate amount of accurate test data always available. Without relevant, useful data, there is no way to test applications to make sure they are operating correctly. 

Although this duty must be a shared one between the application developers and the DBAs, managing and controlling all of the data movement tasks typically falls on the DBAs. And keeping track of what data moved where, when it moved, and why can at times be as much of a challenge as moving the data itself.

Fortunately, there are test data management tools available to not only move the data, but to keep track of it. I’ve written about one of the better Db2 for z/OS data movement tools here in the blog before: Fast and Effective Db2 for z/OS Test Data Management with BCV5. I hope you'll take a moment to click and read that post.

Now BCV5 has been improved with a new reporting feature, to enable users to track the movement of data across their Db2 subsystems. This is a significant new feature that can be used to glean useful information for DBAs, storage administrators, and even by data stewards for data governance.

There are six tables of metadata that BCV5 populates to track the data movement and the copy tasks it performs. These tables are:
  • BCV531.TASK_EXECUTIONS
  • BCV531.TASK_EXECUTIONS_OBJECTS
  • BCV531.TASK_EXECUTIONS_PARAMETERS
  • BCV531.TASK_EXECUTIONS_JOBS
  • BCV531.TASK_EXECUTIONS_RULES
  • BCV531.TASK_EXECUTIONS_MASKING
The information in these tables is updated whenever BCV5 runs a task to copy Db2 data. Users can query these tables just like any other Db2 tables to monitor the details of the BCV5 tasks you have run. This information may be useful for many different IT and business professionals, but let’s take a look at three specific use cases: 
  1. database administration (DBA), 
  2. storage administration, and 
  3. data governance.
DBA tracking
DBAs tasked with moving and refreshing data from one Db2 environment to another are the typical users of BCV5, and therefore they will be one of the primary users of the new reporting tables. Most sites that use BCV5 use it to refresh test data, for example, copying production data to test, or copying unit test data to an integration test set of tables.  

Regardless of the type of data movement that is being undertaken, it is usually being done for multiple tables, tablespaces, and databases. Usually, there will be regularly scheduled processes that copy some of the data, but this is rarely sufficient as there will be on-off requests, special situations, and emergency data refreshes happening all the time. Keeping track of such a hectic morass of copying and refreshing data can be difficult. 

Fortunately, if you are using BCV5 the new Usage Tracker tables can simplify keeping track of data refreshes for DBAs. For example, a DBA looking to find out which BCV5 copy tasks were run during the month of May could code a query like:

 SELECT T.ROWDATE, T.USER, T.TASKNAME 
 FROM BCV531.TASK_EXECUTIONS T 
 WHERE T.ROWDATE BETWEEN ´2021-05-01´ AND ´2021-05-31´ 
 ORDER BY T.ROWDATE ;

This will show all the BCV5 copy tasks that ran during that timeframe, and will look something like this:

ROWDATE      USER     TASKNAME 
---------+---------+---------+---------+--------- 
2021-05-12   USERID1  TSK0001 
2021-05-12   USERID1  TSK0002 
2021-05-20   USERID5  TSKPROD1 
2021-05-21   USERID9  TSKPROD4 
 

The results shown here are just a sample and will likely be a subset of the actual results of running such a query. 

Of course, this is rudimentary information and it is likely that the DBA will want to know more, such as which objects were impacted by these tasks. A query such as the following will come in handy:

SELECT SUBSTR(SRCSCHEMA,1,8) AS SS, 
       SUBSTR(SRCNAME,1,12) AS SN, 
       SUBSTR(TGTSCHEMA,1,8) AS TS, 
       SUBSTR(TGTNAME,1,12) AS TN, 
       SUBSTR(OBJTYPE,1,1), 
       ROWDATE AS DATE_COPIED, 
       SIZEKB 
FROM   BCV531.TASK_EXECUTIONS_OBJECTS 
ORDER BY DATE_COPIED ;

The results here show all the Db2 objects copied by BCV5 showing the source and target names as well as the object type, date copied, and amount of data (in KB) copied:

SS      SN       TS     TN         DATE_COPIED   SIZEKB 
---------+---------+---------+---------+---------+-----
DB500XA TS500X01 DBA001 TS500XA1 S  2021-05-17  1462480 
QUALID  XCL59011 TESTID XCL59011 X  2021-05-17   325040 
QUALID  XCL59012 TESTID XCL59012 X  2021-05-17   125200 
QUALID  XCL59013 TESTID XCL59013 X  2021-05-17   301460 
QUALID  XCL5901C TESTID XCL5901C X  2021-05-17    98400 
QUALID  TEST_TBL TESTID TEST_TBL T  2021-05-17       20 

Again, the results have been truncated as this is intended as an example.

A DBA looking to track down the results of a specific copy task that ran on a specific date might want to run a query like this, to verify which objects were copied. Simply plug in the name of your task and the date it ran:

SELECT T.TASKNAME, O.OBJTYPE, O.SRCSCHEMA, O.SRCNAME, 
       O.TGTSCHEMA, O.TGTNAME, O.PARTITIONS 
FROM   BCV531.TASK_EXECUTIONS         T, 
       BCV531.TASK_EXECUTIONS_OBJECTS O 
WHERE  T.ID = O.EXECID 
AND    T.TASKNAME = ? 
AND    T.ROWDATE = ? ;

Storage Administration Tracking 
Another type of user who might find the Usage Tracker capabilities of BCV5 useful is the storage administrator. Storage administrators are responsible for managing an organization’s disk and tape systems. Additionally, they are also responsible for monitoring storage usage and capacity to ensure that sufficient storage is available for the organization’s IT requirements. 

As such, the storage administrator will likely want to keep an eye on the data movement activities of BCV5. For example, a query such as this one can be used to report on the total amount of data (TS and IX) copied by date:

SELECT ROWDATE AS DATE_COPIED, 
       SUM(SIZEKB) AS TOTAL_KB_COPIED 
FROM   BCV531.TASK_EXECUTIONS_OBJECTS 
WHERE  OBJTYPE IN ('S', 'X') 
GROUP BY ROWDATE ;

Which will return data similar to this:

DATE_COPIED  TOTAL_KB_COPIED 
---------+---------+---------+---------+------ 
2021-06-12         106231270 
2021-06-19         106231270 
2021-06-21        4451457810 
2021-06-26         106231270 
Another potentially useful query, not only for storage administrators and DBAs, but also for application managers, is tracking the amount of actual (tablespace) data copied by date and application. Finding the application name or identifier can be tricky, but if we assume that an application identifier is embedded in the second 2 chars of database name then a query like this can be run:

WITH SIZEBYAPP AS ( 
  SELECT ROWDATE AS DATE_COPIED, 
         SUBSTR(TGTSCHEMA,2,2) AS APPL_NAME, 
         SIZEKB AS SIZE_IN_KB 
  FROM   BCV531.TASK_EXECUTIONS_OBJECTS 
  WHERE OBJTYPE = 'S' 
                  ) 
SELECT DATE_COPIED, APPL_NAME, 
       SUM(SIZE_IN_KB) AS TOTAL_KB_COPIED 
FROM   SIZEBYAPP 
GROUP BY DATE_COPIED, APPL_NAME ;

Which might return a report looking something like this:

DATE_COPIED APPL_NAME TOTAL_KB_COPIED 
---------+------------------+---------+------- 
2021-05-11  EN               46805760 
2021-05-22  BA              242791056 
2021-05-22  BX                4094640 
2021-05-22  CM                1008720 
2021-05-22  DA              270390816 
2021-05-22  OR                  90528 
2021-05-26  PR               55737376 
2021-05-26  XX              537647328 

You can adjust this query if you want to know the amount of index data copied by data and application like so (under the same assumption as above for application identifier):

WITH SIZEBYAPP AS ( 
  SELECT B.ROWDATE AS DATE_COPIED, 
         SUBSTR(T.DBNAME,2,2) AS APPL_NAME, 
         B.SIZEKB AS SIZE_IN_KB 
  FROM   BCV531.TASK_EXECUTIONS_OBJECTS B, 
         SYSIBM.SYSINDEXES              X, 
         SYSIBM.SYSTABLES               T 
 WHERE   B.OBJTYPE = 'X' 
 AND     B.TGTNAME = X.NAME 
 AND     B.TGTSCHEMA = X.CREATOR 
 AND     X.TBNAME = T.NAME 
 AND     X.TBCREATOR = T.CREATOR 
                 ) 
SELECT DATE_COPIED, APPL_NAME, 
       SUM(SIZE_IN_KB) AS TOTAL_KB_COPIED 
FROM   SIZEBYAPP 
GROUP BY DATE_COPIED, APPL_NAME ;

Data Governance Tracking 
Although tracking data movement activities is useful for DBAs, it is also important for data governance reporting. Data governance refers to the processes and standards of ensuring access to high-quality data throughout an organization. Data governance encompasses all aspects of data quality including its accuracy, availability, consistency, integrity, security, and usability. The role of data governance has expanded as data privacy rules and regulations have expanded in response to an increasing number of data breaches and hacker attacks. For example, in early July 2021, Colorado passed the Colorado Privacy Act, meaning that now Colorado, Virginia, and California have passed data privacy legislation that impacts how personal data must be governed. More states are certain to follow their lead… and let’s not forget the European GDPR act!

These type of regulations provide rights for access, deletion, correction, portability, and protection for personally identifiable information, or PII. Note that “portability” is one aspect of data protection covered under the auspices of such regulations… and BCV5 is a mover of data, so you need to be able to track what data moved where, especially when the data that moved contains any PII. 

So, what types of queries can be run using the new BCV5 reporting tables to help satisfy the needs of data governance? 

Well, if you have identified specific tables that have personally identifiable information, and therefore requires specific policies to ensure its privacy and protection, a data steward might want to run a query that shows all of the times that a specific protected table was copied:

SELECT SUBSTR(SRCSCHEMA,1,8) AS SS, 
       SUBSTR(SRCNAME,1,12) AS SN, 
       SUBSTR(TGTSCHEMA,1,8) AS TS, 
       SUBSTR(TGTNAME,1,12) AS TN, 
       ROWDATE AS DATE_COPIED, 
       SIZEKB 
FROM   BCV531.TASK_EXECUTIONS_OBJECTS 
WHERE  OBJTYPE = 'T' 
AND    SRCSCHEMA = ? 
AND    SRCNAME = ? 
ORDER BY DATE_COPIED ;

Simply code the appropriate schema (SRCSCHEMA) and table name (SRCNAME) and this query will show all the times that the particular table (say PROD.CUSTOMERS) was copied. A data governance professional with a list of tables that contains PII could alter this query to accept that list as an IN clause instead of the simple equality clause shown here. 

Additionally, BCV5 can de-identify sensitive data using masking. Whenever a task that requires data to be masked is run, information is captured in the TASK_EXECUTIONS_MASKING table. So, a data governance professional might want to run a query like this one to report on all the masking of sensitive data.

SELECT SUBSTR(TBCREATOR,1,8) AS TBCREATOR,
       SUBSTR(TBNAME,1,12) AS TBNAME, 
       SUBSTR(COLNAME,1,18) AS COLNNAME, 
       METHOD AS MASKING_METHOD, 
       SQLEXPR 
FROM   BCV531.TASK_EXECUTIONS_MASKING 
ORDER BY ROWDATE ;

This can always be modified to join it to the TASK_EXECUTIONS table to obtain the task name if that is important. And with a little manipulation of the query it is possible to look for tables that contain sensitive data that have been copied using BCV5, but have not had masking applied. 

Summary 
BCV5 has offered powerful data movement and masking capabilities for Db2 data for a long time, but now it also offers the ability to track and report on your organization’s Db2 data movement and copy tasks. This new functionality opens a plethora of useful information for BCV5 users.