On March 12, 2010, IBM announced a closed beta for DB2 10 for z/OS and a few select customers are now putting it through its paces. We don’t yet know when DB2 10 will be Generally Available (GA), nor has IBM announced when DB2 Version 8 (in service for six years) will go out of service.
This article offers a DB2 10 for z/OS “Top-10 list,” previewing new functions of a system that offers improved performance with fewer CPUs.
10. Productivity improvements
DB2 10 reduces the need for REORGs by:
- Building compression dictionaries on the fly
- Using index list prefetch, a single SDSNEXIT data set you can share across multiple DB2 subsystems
- Providing the ability to update all statistics stored in the DB2 catalog.
DB2 10 uses a new set of stored procedures and catalog tables to determine what statistics need to be collected and when. It also records what activities are being performed and does it all via an automatic scheduling process. That’s a significant improvement over RUNSTATS, which is challenging to use and can result in collecting unnecessary statistics or not enough statistics, a waste of resources that may not be available.
9. Better access to data
DB2 10 will take advantage of hash access, which should be faster and provide a nice CPU savings. Hashing is a technique that turns a key into a physical location of a row, resulting in one-page access for a matching row on an equal predicate. If you’ve worked with IMS/VS, you may already have some exposure to hash access. If not, no problem. You’ll quickly learn to like it.
There’s also a nice index improvement called index includes. A unique index exists to enforce a unique constraint on a table. Sometimes, a unique index doesn’t contain enough columns in the key to give index-only access. One possible solution is to create two indexes. One index enforces uniqueness and a second index, with additional columns in the key, gives index-only access. DB2 10 will let you "include" additional columns in the index that are not included in the key to enforce uniqueness. One index gives you uniqueness and index-only access.
Both hash access and index include won’t be available until the release of DB2 10 New Function Mode (NFM).
Then there’s access path stability. This is a bit more than what was delivered via an APAR to DB2 9 and V8. Although you still get all the package versioning in DB2 10, you also get: