IDUG NA 2026

In the world of Db2 for z/OS, SQL performance tuning is one of the most critical and often most complex of DBA tasks. While static SQL has long been the standing order for traditional mainframe applications, the growing use of dynamic SQL - driven by modern applications, APIs, and distributed workloads - has changed the tuning game.Dynamic SQL’s agility and flexibility introduces a new layer of complexity that can hinder performance if not well managed and cause inconsistent performance and even be unpredictable. Dynamic SQL is becoming more prevalent as agile development and hybrid workloads become the norm. But with the flexibility of dynamic SQL comes responsibility - DBAs must adapt and develop new ways of monitoring and tuning dynamic workloads in real-time and across the application lifecycle. Colin Oakhill will share a proven approach to bring order to chaos by giving DBAs the visibility, control, and insight they need to ensure that dynamic SQL performs efficiently and predictably. Just like long-time DBAs are used to with static SQL.

Some aspects of Db2 for z/OS BP & GBP technology are plain as day; others, less so. Questions we get from Db2 users worldwide have highlighted for us things that a lot of folks - even very experienced Db2 people - get wrong. In this session we'll try to bring clarity to oft-misunderstood areas of Db2 for z/OS BP & GBP technology.

This presentation will demonstrate how to find, analyze and tune the top 40 expensive queries. Special emphasis will be on finding the cursors that could benefit from multi-row fetch. In addition, discussions on designing perfect indexes, MQTs and AQTs, along with manual query rewrite, SQL exploitation and most important, proper use.

The AI problems landing on the desks of DBAs and application developers this year don't all need the same model. Some call for a Large Language Model. Others need a forecasting model, a clustering algorithm, or an embedding model. Plenty can be solved by predictive techniques that have been working quietly for years. One drawer of the toolbox has been getting all the attention. The skill that matters most, knowing which tool fits the problem, has gone quiet. If your data lives in Db2, much of the toolbox is already there.This session lays the toolbox out on the table. Two axes, what data you have and whether you have labels, organize the AI landscape into a 2×2 you can hold in your head: predictive and unsupervised models for structured data, the same for unstructured data, plus embeddings and generative AI.Then we map every drawer to a Db2 capability you can use today. In-database machine learning for classification, regression, and clustering. Native vector storage and similarity search. Built-in integration with external embedding and language models. Python frameworks for retrieval-augmented applications and AI agents. We close with production patterns we've seen ship, across the full range of model types and use cases.Whether you live in SQL or Python, you'll leave knowing which tool to reach for, and where to start.