Take a Sneak Peek at the New DB2 11 for z/OS
In last month’s column, IDUG president Joe Borges outlined his vision for IDUG in the organization’s 25th year. But IDUG isn’t alone in celebrating 2013; this year also marks 30 years since IBM® DB2® for z/OS® became generally available. In an industry as fast-moving as ours, it’s an amazing achievement for a product to be around for so long and still remain as innovative and relevant as the day it was first released. It’s fitting, then, that IBM has chosen to begin the early support program (ESP) for the latest release of DB2 at this auspicious time.
As part of the ESP for DB2 11 (previously known as Sequoia), customers from around the world are helping to test and evaluate the new release. In this article, I’ll briefly examine some of the planned new features that the DB2 community can look forward to—but remember that the exact content of the release will depend on the feedback IBM receives from customers during testing. We’ll have to wait until the product is formally announced before we can be certain that all of the features below make it into the release in the form that I’ve described. But it’s certainly worth an early look.
What’s new in DB2 11?
The new release of DB2 for z/OS contains a great mix of performance, availability and productivity features. Here’s a quick rundown of the ones that I think are going to be most popular and interesting:
- Application CPU savings. DB2 10 delivered some very significant CPU savings for many customers, and will be a very tough act for IBM to follow. It’s still too early to quantify the savings we can expect from a typical DB2 11 upgrade—and any general figures quoted have so many caveats and assumptions associated with them they are essentially meaningless in terms of what a given customer can expect to see). However, if your production workload includes significant amounts of heavy INSERT work, SAP applications, or some specific types of query processing, you can anticipate substantial CPU savings.Many customers make extensive use of DB2 compression to reduce direct access storage device (DASD) space and improve buffer pool performance, and DB2’s decompression algorithms have also been optimized to reduce CPU for applications reading compressed tables. Overall, most customers should see at least some CPU improvements as a result of moving to DB2 11—in contrast to the 5-10 percent CPU regression that was typical before DB2 10.
- Data sharing performance. Despite being one of the unique features offered by DB2 for z/OS, data sharing has seen relatively little enhancement during the past few releases. That changes in DB2 11, with a comprehensive set of performance tweaks to improve buffer pool management, lock management and restart/recovery, and to reduce CPU overhead due to log record sequence number (LRSN) spin.
- Transparent data archiving. Many applications include provision for archiving older data at the design phase, but these features are often among the first to be shelved if project timescales start to slip. New features in DB2 11 build on the temporal data support introduced in DB2 10, and give developers an efficient and transparent way to access both current and archived data within a single SQL statement.
- Utility enhancements. The RUNSTATs and LOAD utilities have been enhanced to make greater use of any available zIIP Specialty Engines, potentially improving performance and reducing CPU consumption. Online REORG has had a major overhaul, with improved availability, support for automatic mapping tables and the ability to handle partition rebalancing. At the same time, the requirement to run REORG at all has been reduced by enabling DB2 to automatically clean up pseudo-deleted index entries. It’s worth noting that, while many utility enhancements were available in Conversion Mode (CM) in previous releases, the majority of the DB2 11 enhancements are not accessible until you have entered New Function Mode (NFM).
- Log scalability. Some customers are approaching (or have already hit) the maximum log relative byte address (RBA) range possible using the existing 6-byte format. DB2 11 expands RBA/LRSN values to 10 bytes, increasing the addressable range to a mind-boggling 1 yottabyte (280).
- Online schema change. IBM is steadily working through the list of most-requested features to allow database objects to be changed while maintaining availability. New in DB2 11 is the capability to DROP columns from a table, and alter partitioning limit keys. Point-in-time recovery is also now supported where there are deferred schema changes.
- Application compatibility. One of the most common obstacles that prevents customers from migrating to a new release is the need to implement any code changes required due to application incompatibilities such as new reserved words. A new application compatibility feature in DB2 11 allows such applications to continue to run under the new release until those changes can be implemented, removing them as an upgrade dependency.
- zEC12 hardware exploitation. The latest generation of IBM zEnterprise® EC12 servers—announced in August 2012—include a new feature known as Flash Express. This is a special solid-state disk (SSD) array that acts as a new memory tier between the redundant array of independent memory (RAIM) and disk. DB2 11 will be able to use this as a better alternative to auxiliary/disk if there is insufficient real storage available (although the best advice is still to not over-commit your real storage). Customers will also be able to use 1 MB page frames without having to page-fix their buffer pools (DB2 10 mandated page-fixing if large pages were used) and page frames of up to 2GB will be supported.
Preparing for a DB2 11 upgrade
If you’re tempted by these new features and want to begin positioning your environment for the upgrade to DB2 11, IBM has already announced two key prerequisites: you must be on a System z10 server or later (EC12 is required for exploitation of the new Flash Express feature described above), and running z/OS V1.13 or later. Perhaps more importantly, you will need to be running DB2 10 for z/OS in NFM before you can start the upgrade; this means IBM is adhering to the normal practice of only supporting upgrades from the previous release, with no “skip release” functionality being offered this time around. If you’re still running DB2 9 for z/OS, you’re probably already aware that IBM has announced that it will be ending support for that release in June 2014. So your plans for migration to DB2 10 should be in place; if not, now is a good time to start.
One other significant item to consider is package validity. Many customers used to avoid REBINDing their DB2 packages due to fear of performance regression, so DB2 had to continue to support old internal package structures from many releases. The availability of the plan management feature in DB2 9, together with associated BIND parameters such as APREUSE and APCOMPARE, allows DBAs and developers to better manage the REBIND process, thereby removing the major reason for avoiding REBINDs at least once per release. As of DB2 11, only package structures from a currently supported release of DB2 will be tolerated—so customers with packages bound before DB2 9 will have to rebind these packages as part of their DB2 11 upgrade preparation (or have DB2 do it automatically when they are first referenced).
Learn more at IDUG conferences
I hope this article has given you a taste of at least some of the great new features that will soon be heading our way as part of DB2 11 for z/OS. Expect to hear a lot more as we draw nearer to general availability: all of this year’s IDUG Technical Conferences will include one or more DB2 11 sessions, and I’m about to begin working with several of my IDUG volunteer colleagues to produce an in-depth DB2 11 technical white paper. More news on that soon.
In the meantime, let me know what you think about the advances we anticipate in DB2 11 for z/OS. Please share any thoughts or questions in the comments.