IBM Db2 for z/OS

Db2 and AI - The future is now!

By Paul McWilliams posted Mon December 09, 2019 03:08 PM


This entry was originally published on 2018-09-21.

IBM Db2ZAI is a new product that empowers the optimizer in the Db2 for z/OS engine to determine the best-performing query access paths, based on workload characteristics.

We asked members of the Db2 development team for their thoughts on what IBM Db2 AI for z/OS will do for our clients, both now and in the future, and here's what they said*:

Mo Townsend, Director of Development, Db2 for z/OS

“The first version is focused on delivering a value proposition of improved performance or reduced CPU. We will be looking to add additional capability in terms of optimizer refinement in future releases.

Our vision is to expand the value proposition into other areas as well. One of these additional value themes is automation of selected tasks currently performed by DBAs and system programmers. Where Db2 has internal insight into the setting of certain parameters, for example, or the execution of certain tasks, we would like to have Db2 do those things rather than have DBAs have to spend their time doing that, so the talent can invest in the things that are going to deliver the highest value.

The third value area we see to date is risk management or risk reduction. One example we have in mind is looking at the client's Db2 system and assessing which HIPERS or high priority APARs they are missing. We have a tool that does this today, but what we want to be able to provide in the future, is to leverage ML to assess which capabilities they are using in their system so we can predict which HIPERS they are most likely to encounter, then recommend for which ones they should prioritize for installation.

As exciting as this first release is, there's lots more innovation opportunity in this marriage between Db2 for z/OS and AI capabilities!”

Tom Beavin, DB2 for z/OS AI Development team

“Db2ZAI is going to provide benefits to our clients in terms of query performance, related to being able to choose better access paths. So we are focused on helping the optimizer with two challenges that it has today: 1) understanding if there is a pattern to the host var values that are being seen by the query and 2) if there is a pattern to the cursor behavior in the application that is driving a particular query, and that is particularly challenging to the optimizer right now because we don't collect statistics on those things today. So, this is going to allow Db2 to observe that behavior and take appropriate action when it sees a pattern there, relieving the burden on the application programmer or DBA for Db2.

So, that is kind of the initial benefit that Db2ZAI is bringing. Its real promise is in what we are able to do down the road ... so, bringing machine learning into Db2 allows us to make Db2 self-managing, self-tuning. To do that, which has been a long-term goal, the challenge is how do you make it so that Db2 can be just as good or better at managing, as it were, turning those knobs, adjusting the different parameters, as a DBA. How can you get the skill, the level of performance built into Db2 that you could with a skilled DBA? Where I think the breakthrough is, is with machine learning and the ability to create models; it provides us an avenue now to do that, that was not there before.

And so, we are starting down that path, with this initial release of Db2ZAI, that focuses on certain aspects of query performance. There is a clear benefit there. But there are so many more things that excite me about this product. Db2ZAI allows us to solve so many long-standing problems or wish-list items in Db2, and with the end result being that we have a database that gives the user fantastic query performance, out of the box. There might be some learning time now, with this approach, but after the learning period, Db2 is a highly-tuned system that is adapting to the workload that is being run, and is able to keep track of what's going on in the system and adapt as it needs to. That is pretty cool!”

Vince Rabsatt, DB2 for z/OS AI Development team

“Db2ZAI provides the ability to tailor Db2 to a specific customer, and the data that is most relevant to that customer, by having the ability to process the data that the customer uses more frequently, and learning from the data patterns and making optimizations around that. Rather than being general to a broad range of customers, it's more customer specific. An analogy would be, instead of going into a department store and buying a suit, you have a tailor coming to your house and making a suit specifically for you by measuring you. It's basically about learning what is important to you and customizing the database specifically for you.”

For the future, we can see the benefits, as things are constantly changing so that when our customers' needs change, Db2ZAI will be able to learn those changes and adapt. So I see the benefit of adaptability as time and data change, Db2 will be able to change with it. Optimization is not a place where we would want to stop; it's definitely where we would want to start and branch out and touch other areas.”

* The above statements reflect the opinions of those interviewed, and do not constitute a guarantee of future product capabilities.