IDUG EMEA 2024

This lab session is provide the attendees the hands-on experience to explore how AI enabled monitoring solution, with a trained ML model, can help manage DB2 and z/OS performance to empower you to become more proactive before the problem damages the system.  The lab is hosted on an IBM live z/OS environment via internet.  The instructor will be on-site providing a lab guide and answering questions.  The lab guide gives step-by-step instructions to take you through Db2 problem solving scenarios with AI-ML scoring in action.  Bring your own laptop device, and we will set it up for you to login to this AI-ML lab on a live IBM Z system.

Wouldn't it be great if Db2 could automatically determine ideal access plans based on hidden correlations discovered in your table data? That's the idea behind Db2's AI Query Optimizer, a key new feature in Db2 v12 for LUW. This talk will go through the basics of how the Db2 AI Query Optimizer automatically trains machine learning models that discover hidden data correlations. These machine learning models are later used at SQL query compilation time to generate better runtime access plans, making your queries run faster. All of this is done automatically, under the covers.

Join me on a journey through our first year of deploying SQL Data Insights (SDI) into production. Discover how SDI impacted our business processes, changed stakeholder perceptions of the mainframe, and enhanced our approach to data analysis.
We will start with a review of our first use case and discuss the resources consumed for training and scoring, as well as the impact SDI had on our business processes. 
SDI's most evident benefit is its ability to analyze data where it sits without having to export, transfer and reload it, and with minimal data science dependency. But how does it compare to more traditional approaches? To answer that, we'll compare SDI with other machine learning algorithms by applying them to a frequently used dataset for predicting diamond prices. We'll delve into the accuracy of the algorithms and see how SDI stacks up against data scientists' most loved algorithms.
Following that, we'll explore a brand-new use case that applies the new commonality-function in Db2 FL504. We'll discuss its usefulness, resource requirements, and potential impact on business processes, providing more practical insights into the power of SDI.
And finally, I'll share a use case that demonstrates how to identify outliers using SDI with Db2 monitoring data, ready to be applied in your own shop. 
Don't miss this opportunity to gain practical insights into the power of SDI and how it can transform your data analysis!

The Db2 AI Query Optimizer builds, trains, and uses Deep Learning Neural Networks to provide advanced query optimization capabilities. Introducing deep learning safely onto production database management systems is a monumental task. Learn about the tremendous engineering challenges we confronted as we transitioned the Db2 ML Optimizer Tech Preview into a production-ready feature in v12. Some of the many challenges faced were how to deliver good machine learning model accuracy across a wide variety of use cases, how to minimize impact to existing workloads running on the database management system, and how to minimize machine learning model storage requirements. Knowing how the Db2 AI Query Optimizer was designed and how these engineering challenges were tackled will give you the confidence needed to try out this new feature on your own database systems.

In this session you will find Isbank's experience with Db2 AI for z/OS. It will cover all POC and production experiences for the tested features of Db2 AI for z/OS as well as details about implementation project. What ill be your benefits if you decide to use Db2 AI in your shop? What does Db2 AI offer in a system assessment, SQL optimization and distributed data control parts? What are the new features that we are looking forward to in new releases? You will find all the answers in this presentation.

This session introduces the Lakehouse concept and what sets it apart from data lakes and traditional data warehouses. We focus on mainframe operational data stores as data feeds, and how such feeds can be integrated with other, non-mainframe data sources, how Lakehouses re-invent analytics and how machine learning training pipelines can be fed using data extracted and refined on a Lakehouse. 
We will use watsonx.data as the Lakehouse implementation, but the approach translates to any Lakehouse because of the use of open data formats.