2025-03-03

Parallel workers without thrashing your NVMe budget

Priya Desai · performance · hardware

Query Planner Deep Dive attendees capture iostat samples while toggling max_parallel_workers_per_gather. The visceral charts convince stakeholders faster than abstract warnings.

We also map temperature curves on lab laptops so participants respect thermal throttling during long benchmarks. Realistic hardware stories beat synthetic cloud-only labs.

Capstone presentations must include a rollback plan if production decides to disable parallel query globally. Planning for failure is part of the grade.