Conversation about Autonomous PostgreSQL: Luigi Nardi, CEO, DBtune

I recently sat down with Luigi Nardi, founder and CEO of DBtune, to discuss a subject close to my heart: the evolution of autonomous PostgreSQL. It was an incredibly insightful conversation that traced the path from deep academic research to real-world production breakthroughs.

Watch the full interview above — here are a few of my notes and key takeaways.

The technology behind DBtune originated at Stanford University in 2017, where Luigi was developing probabilistic machine learning models to optimize computer systems. Recognizing that database systems were a perfect fit for this technology, Luigi built the first software prototype, eventually incorporating DBtune in 2020 with crucial guidance from PostgreSQL community leaders like Magnus Hagander.

While traditional rule-based tools like PGTune have served the community well, Luigi explains that they are limited by a "template-based" approach. Modern production environments are too complex for static rules; they require AI-assisted engineering to navigate vast solution spaces and recognize intricate workload patterns.

• Multi-dimensional optimization: Unlike a human DBA who might tune one parameter at a time, AI can simultaneously adjust interconnected variables — such as disk cost hints to the planner and parallel workers — to achieve a multiplier effect on performance.
• Beyond managed services: Even in the age of AWS RDS and Azure PostgreSQL, tuning remains vital. Cloud providers typically use predefined parameters based on instance size rather than the actual workload. DBtune has demonstrated that autonomous tuning can achieve a 10x improvement in performance and a 50% reduction in CPU usage in real-world production environments.

Next, Luigi draws a compelling parallel between databases and the decades-long evolution of self-driving cars. While the vision of a completely autonomous PostgreSQL is bold, DBtune is already executing a roadmap to make it a reality. We concluded our interview with a discussion on DBtune's roadmap and upcoming features.

• Expanding automation: The next phase for DBtune includes automating classical database challenges such as vacuuming, indexing, and session-level tuning.
• Modern mechanisms: The company is also focused on optimizing newer PostgreSQL features, such as asynchronous I/O, to ensure the entire stack is running at peak efficiency.

Luigi's message is clear: the agentic AI-based database tuning technology is ready for prime time. By automating the "tedious process" of tuning, DBtune allows developers and DBAs to stop worrying about tuning their instances, focus on building the future, and create significantly more efficient solutions.