- DBPROTCL is no longer supported; the DBPROTOCOL bind option is DRDA by default.
- MAX_OPT_ELAP specifies the maximum amount of elapsed time the DB2 optimizer can consume.
- MORE_UNION_DISTRIBUTION when set ON, can improve performance of queries using views defined with UNION ALL.
- RELCURHL is an option to hold a lock over a commit.
- STORPROC is the parameter for creating stored procedures.
- SUPPRESS_TS_CONV_WARNING is the option to turn off messages when DB2 converts a table space from index-controlled to table-controlled partitioning.
- TABLES_JOINED_THRESHOLD sets a limit (16) on table joins.
These ZPARMs would have been removed in a DB2 V8 to DB2 9 upgrade.
It’s important to be aware of which subsystem parameter default settings were changed in DB2 10. A failure to specify different options in the DSNZPARM parameters in your Assembler job means you accept the default settings. Consider whether there are benefits to using a different value. To facilitate your planning efforts, we’ve provided the default changes. See Figure 1 if you’re planning a skip-level migration from DB2 V8 to DB2 10. Refer to Figure 2 if you’re moving from DB2 9 to DB2 10.
You should also be aware of changes to maximum values for DSNZPARMs. For example, the big winners in DB2 10 are DSNZPARMs that deal with threads. The maximum number of user threads (CTHREAD), maximum batch threads (IDBACK), maximum foreground threads (IDFORE), and maximum DBATs or distributed threads (MAXDBAT) increased tenfold; CTHREAD, IDFORE, and IDBACK increased from 2,000 to 20,000, and MAXDBAT increased from 1,999 to 19,999.
These four increases are significant because they could reduce the dependency on data sharing. To achieve higher thread counts in previous releases, some users had to take advantage of data sharing. Those users may now be able to reconsider that decision.
MAXOFILR is another DSNZPARM that made a dramatic jump in size; the maximum number of data sets that can be open concurrently for processing Large Object (LOB) file references increased from 2,000 to 20,000.
DSSTIME, the interval in minutes DB2 waits before resetting the data set statistics for online performance monitors, dropped from a maximum of 1,440 to 60 minutes.
CACHEPAC and CACHERAC, the authorization caches for packages and routines, both increased from 5MB to 10MB, essentially doubling in size.
STATIME is the time interval in minutes between statistic records collection. The maximum value dropped from 1,440 to 60 minutes. That’s still a pretty large number, but the default was lowered to one. That’s good news because when statistics are collected at longer time intervals, you can’t report on any interval less than the interval used for collection. Sometimes, statistics collected at a granular interval are necessary to diagnose a performance problem.
STATIME and SYNCVAL, the interval used to sync up your data sharing members, in DB2 10 only apply to IFCIDs 0105, 0106, and 0199. IFCIDs 0001, 0002, 0202, 0217, 0225, and 0230 are now always written at a fixed, one-minute interval; they’re no longer affected by the STATIME value specified.
MONSIZE, the default buffer size used for monitor traces, was increased from 16MB to 64MB. The benefits should be apparent.
This article doesn’t cover the many new DSNZPARM keywords added in DB2 10; that will be the basis for a future article. Until then, enjoy your powerful DSNZPARMs. To learn more, visit the DB2 for z/OS Information Center at http://publib.boulder.ibm.com/infocenter/dzichelp/v2r2/index.jsp?topic=/com.ibm.db2z10.doc/db2z_10_prodhome.htm.