No video of the event yet, sorry!

This training session will consist of 2 parts. The first is a short presentation that introduces the concept and the basics of database parameter tuning, why it’s so important and the business value that can be realized when systems and applications are optimally tuned. We will explore existing options for addressing this critical task, including manual tuning and existing open-source solutions. In the second part we will provide a practical, hands-on tutorial session that covers utilizing a fully automated solution on a synthetic workload and offers a straightforward resolution to this challenge.

Database management systems (DBMS) expose configurable parameters that control their runtime behavior, such as shared_buffers and work_mem. The general trend is that over the years, new parameters are added in each new system release, resulting in an explosion of parameters — PostgreSQL has hundreds of parameters. Setting some of these parameters correctly improves the application performance and saves substantial cloud resources. Given the complex interaction of these parameters, tuning them requires considerable effort. The available approaches to date are based on manual tuning and open-source DIY tools. Manual tuning involves the “dark art” of analyzing various metrics, such as query response times, CPU usage, disk usage, and memory utilization, to identify performance bottlenecks. However, it is slow and requires a high level of expertise. In contrast, open-source tools, such as PGTune, often provide a one-size-fits-all solution, and their rule-of-thumb approach may not be effective for all workloads. These tools are often used as a starting point for manual tuning.

We take a radically different approach and introduce DBtune, a fully automated parameter tuning service that is both workload and machine resource aware. The optimizer is based on machine learning technology that customizes each optimization to the specific customer workload and hardware in use. DBtune observes database key performance indicators and trains recommendation models to select parameters that lead to improved performance. This approach streamlines and scales the database tuning process across a potentially unlimited number of instances, enhancing efficiency. By leveraging DBtune, users can significantly enhance their database performance while concurrently reducing cloud expenses and the environmental impact of their database operations. DBtune's fully automated tuning solution is not open source.

Speaker bio

Luigi Nardi is the founder and CEO of DBtune (www.dbtune.com), a leading company driving advancements in AI, database systems, and cloud computing. Previously an assistant professor of machine learning at Lund University and a research staff at Stanford University, Luigi's expertise centers around black-box optimization theory and practice. Luigi's journey includes a post-doctoral position at Imperial College London and a role as a software engineer at Murex S.A.S., following his Ph.D. program in applied mathematics at Université Pierre et Marie Curie in Paris in 2011.