Friday, June 29, 2007

Implicitly Created Database Objects [DB2 9 for z/OS]

Today we continue our series on new features in DB2 V9 with a quick discussion of implicitly created database objects. To understand what this is, let’s first review the way DB2 works today (pre-V9). If you issue a CREATE TABLE statement and do not specify the database and table space into which the table is to be created, DB2 will automagically create a new table space in the default database (DSNDB04).

Not being content with that, DB2 9 extends this capability with the ability to implicitly create additional types of database objects. By coding your CREATE TABLE statement with the proper options you can implicitly create any or all of the following:
  • Database
  • Table space
  • Index to enforce Primary Key uniqueness
  • Index to enforce unique constraint
  • ROWID index (if the ROWID is defined as GENERATED BY DEFAULT)
  • LOB structures (LOB table space, auxiliary table, auxiliary index)

  • OK, so how does this happen? Let’s go down the list. If you fail to specify the IN clause on a CREATE TABLE, DB2 works a bit differently. In the past, DB2 would simply create an implicit table space in DSNDB04. As of DB2 9, the database is involved as well. DB2 will either create an implicit database or use a previously implicitly created database. The names of these implicitly created databases will range from DSN00001 to DSN60000. The first time, DB2 will create DSN00001, the second DSN00002, and so on until we reach DSN60000. The next time, DB2 will wrap around and start again from the beginning, using existing implicitly created databases. For the implicitly created databases, the STOGROUP will be set to SYSDEFLT; buffer pool values are determined via DSNZPARMs.

    Next up is the table space. Although DB2 has supported implicitly created table spaces forever, there are some twists in DB2 9. First of all, you cannot create simple table spaces any longer, so all implicitly created table spaces will be segmented. In compatibility mode (CM), a implicitly created table spaces will be defined as SEGSIZE 4 and LOCKSIZE ROW. After migrating to new function mode (NFM) your implicitly created table spaces will be created as partition by growth table spaces. The options uses will be SEGSIZE 4, DSSIZE 4G, MAXPARTITIONS 256, LOCKSIZE ROW, and LOCKMAX SYSTEM.

    As for the rest of the objects in the list, these system-required objects will always be implicitly created if the table space is created implicitly. For indexes that support the primary key or unique constraints, the names will be generated using a combination of the table name and randomly generated characters.

    OK, so now that you know about the ability of DB2 9 to implicitly create objects, let me give you some advice. Whenever possible, don’t rely on it. It is much better, if at all feasible, for your DBAs to explicitly create and name all database objects as needed. Yes, it takes more time, but it gives you more control. You can explicitly decide which objects go into which database; you can explicitly set paramters; etc.

    So, this is a nice new feature and it can enable DB2 to do some definitional things for you automatically. But most DBAs will want to continue to do things the traditional way, that is, building their DDL themselves without relying on implicitly

    Monday, June 18, 2007

    Skipping Locked Rows [DB2 9 for z/OS]

    In DB2 9 it is possible for a transaction to skip over rows that are locked. This can be accomplished by means of the SKIP LOCKED DATA option within your SQL statement(s). SKIP LOCKED DATA can be specified in SELECT, SELECT INTO, and PREPARE, as well as searched UPDATE and DELETE statements. You can also use the SKIP LOCKED DATA option with the UNLOAD utility.

    Of course, if a program skips over locked data then that data is not accessed and the program will not have it available. When this option is used DB2 will just skip over any locked data instead of waiting for it to be unlocked. The benefit, of course, is improved performance because you will not incur any lock wait time. But it comes at the cost of not accessing the locked data at all. This means that you should only utilize this clause when your program can tolerate skipping over some data.

    The SKIP LOCKED DATA option is compatible with cursor stability (CS) isolation and read stability (RS) isolation. But it cannot be used with uncommitted read (UR) or repeatable read (RR) isolation levels. DB2 will simply ignore the SKIP LOCKED DATA clause under UR and RR isolation levels.

    Additionally, SKIP LOCKED DATA works only with row locks and page locks. That means that SKIP LOCKED DATA does not apply to table, partition, LOB, XML, or table space locks.
    Let's look at an example. Suppose we have a table with 5 rows in it that looks like this:

    --- ------ -------
    1 JOE MAMA

    Assume row level locking. Next assume that an UPDATE statement is run against the table changing FNAME to JIM WHERE LNAME = 'KIM'. And it is hanging out there without a COMMIT. Next, we run:


    The count returned would be 3 because DB2 skips the two locked rows (rows 3 and 5). And, of course, if the locks are released the count would be 5 again.

    Wednesday, June 13, 2007

    Do You Want to Ignore Clustering? [DB2 9 for z/OS]

    DB2 9 for z/OS offers a new DDL parameter for your tables: APPEND. If you specify APPEND NO, which is the default, DB2 will operate as you are accustomed to it operating. That is, when rows are inserted or loaded DB2 will attempt to sequence them based on the clustering index key.

    If you specify APPEND YES though, DB2 will ignore clustering during inserts and online LOAD processing. Instead of attempting to maintain clustering, DB2 will just append the rows at the end of the table or partition. If the table space is partition-by-growth (new DB2 9 feature) then DB2 can use any partition with space available at the end; for range-partitioned table spaces, obviously DB2 will append the data to the end of the partition corresponding to the range for the value(s) being inserted.

    You might want to choose this option to speed up the addition of new data. Appending data is faster because DB2 does not have to search for the proper place to maintain clustering. And you can always re-cluster the table by running a REORG.

    The APPEND option cannot be specified on LOB tables, XML tables, or tables in work files.

    To track the state of the APPEND option there is a new column, APPEND, in the DB2 Catalog in SYSTABLES. Its value will be either ‘Y’ or ‘N’.

    Monday, June 04, 2007

    Reordered Row Format [DB2 9 for z/OS]

    If you’ve worked with DB2 for awhile, especially as a DBA, you’ve probably heard the advice to re-arrange the columns of your tables to optimize logging efficiency. Basically, the more data that DB2 has to log, the more overhead your programs will incur, and performance will degrade. DB2 will log data from the first byte changed to the last byte changed – unless the row is variable, in which case DB2 will log from the first byte changed to the end of the row – unless the change does not cause the length of the variable row to change, in which case DB2 goes back to logging from the first byte changed to the last byte changed.

    So, the advice goes something like this: put you static columns (those that do not change frequently) at the beginning of the row and your dynamic columns (those that will change more frequently) at the end of the row. And put your variable columns at the end of each. This would make your row look something like this:

    [Static fixed-length cols]
    [Static variable cols]
    [Dynamic fixed-length cols]
    [Dynamic variable cols]

    Make sense?

    Well, DB2 9 for z/OS takes this advice to heart (sort of). In New Function Mode (NFM), for new table spaces, DB2 will automatically put the variable columns at the end of the row. This is called reordered row format (RRF); the row format we are all familiar with today is now referred to as basic row format (BRF). This is all how the row is stored – it does not mean that your DDL is changed nor does it require changes to anything external or how you access the rows.

    To summarize, this means that a row in RRF will store the fixed-length columns first and the variable columns at the end. Pointers within the row will point to the beginning of the variable columns.

    So far so good, right? Well, we DB2 will also convert our old table spaces to RRF over time. Once we are in DB2 9 NFM, a REORG or a LOAD REPLACE will cause a change from BRF to RRF. So run a LOAD REPLACE a table space in NFM and the row format changes. REORG a partition and the row format for that partition changes. And yes, you can have a partitioned table space with some partitions in BRF and some in RRF.

    With BRF we can be sure that DB2 is putting our variable columns at the end of the row – where they belong. But it still is not helping us with placing static columns before the dynamic ones. You’ll still have to guide DB2 to do that.