IDUG EMEA 2026

A critical Db2 for z/OS purge batch is repeatedly abending after hitting CPU limits—without deleting a single row. What followed was a six-month troubleshooting marathon. With the batch spending over 80% of its time in I/O wait and racking up 500 million GETPAGEs on a 2-million-row table, standard tuning simply wasn't enough.This session dives into the gritty reality of complex performance tuning: the dead ends, the false starts, and the eventual breakthroughs. We will walk through the entire investigation—from utilizing monitoring tools to experimenting with UTS conversions and dedicated buffer pools. Finally, we’ll reveal how rethinking our DELETE index strategy, host variables, and leveraging ON DELETE CASCADE NOT ENFORCED transformed a CPU-crushing failure into a consistent, 20-minute purge process. Attendees will leave with a practical, battle-tested methodology for rescuing failing batch processes.