DB2 10 Technical Overview at IOD Conference

Today I attended the IOD conference and had the opportunity to listen to Jeff Josten present an technical overview of DB2 10 for z/OS. Even though information and specifications have been trickling out on DB2 10 for z/OS over the course of the past year or so, this is the first DB2 10 presentation I have attended subsequent to the GA announcement IBM made last week (announced October 19th, General Availability on October 22nd, 2010). So I’m fairly certain that everything Jeff will talk about will be officially part of DB2 10, instead of just the rumors, hints and allegations proffered prior to the GA announcement. I don’t think anything to be covered will surprise me, but we’ll see.

Jeff started off saying that the technical strategy for DB2 10 was four-pronged:

Continuous availability

Performance and scalability

Ease of management

Advance application features

One of the key value propositions is the deep synergy with the System z hardware. Because the code does not need to be ported all over the place, it can take advantage of the hardware capabilities that bring improved performance and efficiency.

DB2 10 is being promoted as delivering 5 to 10 percent CPU batch and transaction performance improvement out-of-the-box; 20 percent for new workloads. And then an additional 10 percent when you start using new features. This seems to be the high-level talking point for DB2 10 – but that is okay, it is a very good one!

DB2 V8 will go out of service April 2012. But with the ability to go staright from V8 to 10, I’m betting a lot of V8 shops will skip 9 altogether and go right to 10. But it looks like you have just under 2 years to get off of V8, though.

DB2 10 takes advantage of many zEnterprise (the new mainframe announced a couple of months ago) features to deliver scalability. Examples include improved compression, cache optimization, blades for running the Smart Analytics Optimizer, etc.

Jeff mentioned that we’ll be able to support 10 times more users by avoiding memory constraints in DB2 10. That is a big scalability improvement!

DB2 10 went through the largest beta ever: 23 customers and more than 80 vendors. The focus was on testing production level workloads to ensure that the release is stable. That is different than past betas where the focus was on testing new features. And 22 of the beta customers are planning to go into production with DB2 10 next year. Impressive!

Before diving into the technical details, Jeff mentioned that not much has changed versus what IBM has been talking about over the past months. A couple late add features include hash performance, BIND performance, REBIND not required for packages flagged as private protocol (but they will fail if they actually use private protocol), and a new ZPARM for default SEGSIZE for DDL compatibility.

Post GA delivery items include APREUSE and APCOMPARE (for reusing access paths instead of using “hints”) because beta testing exposed some quality items, the ability to delete a data sharing member, inline LOBs for SPT01, online REORG concurrency for materializing deferred ALTERs, and some temporal enhancements (e.g. TIMESTAMP WITH TIMEZONE support).

High performance DBATs (DDF threads) were forced to be RELEASE COMMIT. This gave good storage usage but at the expense of CPU for releasing resources at COMMIT and putting the thread back on the queue. For DB2 10, customers can use RELEASE DEALLOCATE which will keep the thread assigned to the distributed requestor so the next time he comes in he can reuse the thread. This is a nice feature for shops with heavy distributed usage.

Another nice thing right out of the box is parallel index updating at INSERT. This used to be synchronous, each index modified one after the other. Now, the indexes can be modified in parallel, which should be a nice performance improvement for many shops!

There is also a new buffer pool option for a “fully in memory” object for reading the data and pinning it in buffers. Don’t know about you, but I’ve been wanting that for years.

Things that require a REBIND will include most access path enhancements, query parallelism improvements, IN list performance improvements, and Stage 2 predicates being pushed to Stage 1.

Things that require NFM include DB2 Catalog concurrency, compress on insert capability, most utility enhancements, LOB streaming between DDF and the rest of DB2, INSERT improvements for universal table spaces, faster FETCH and INSERT with lower virtual storage consumption, SQL Procedure Language performance improvements, efficient caching of dynamic SQL with literals, as well as a few other “things.”

And then there are things that require NFM and DBA work, such as hashing, index include columns, inline LOBs, DEFINNE NO for LOB and XML columns, MEMBER CLUSTER for universal table spaces, and online REORG for all DB2 Catalog and Directory table spaces.

So it looks like our favorite DBMS is continuing to grow and expand offering high performance functionality and features that are unparalleled in the industry. Indeed, there is a lot of great and exciting new “stuff” on the way in DB2 10.