Keynotes

Query Optimization in the Wild: From Cross-platform Data Systems to Multi-Agent Systems

Query optimization is (or should be) at the core of any data management system. Without effective and efficient query optimization, even a sophisticated data system is bound to underperform. In this talk, I will first share our query optimization journey in Apache Wayang, an open source framework designed to unify analytics across diverse data sources and data processing engines. I will begin with traditional cost-based optimization and show its limitations in cross-platform settings. I will then briefly discuss learning-based approaches and explain why, despite their promise, they still struggle when optimization depends on enumerating a huge search space of plans. Motivated by these limitations, I will then introduce a different perspective on query optimization: a generate-and-explore approach that replaces the traditional enumerate-score-prune methods with generative models and an exploration-driven feedback loop. I will conclude my talk by motivating the need for query optimization in multi-agent systems, an emerging class of data systems in which multiple agents must coordinate, reason, and adapt to deliver meaningful insights.

Zoi Kaoudi is an Associate Professor at the IT University of Copenhagen. Her research interests lie in the intersection of machine learning systems, data management, and knowledge graphs. She has previously worked as a Senior Researcher at the Technical University of Berlin, as a Scientist at the Qatar Computing Research Institute (QCRI), as a visiting researcher at IMIS-Athena Research Center, and as a postdoctoral researcher at Inria Saclay. She received her Ph.D. from the National and Kapodistrian University of Athens in 2011. She is currently a Proceedings and Metadata Chair of ISWC 2023. Previously, she has been an Associate Editor of SIGMOD 2022, proceedings chair of EDBT 2019, co-chair of the TKDE poster track co-located with ICDE 2018, and co-organizer of the MLDAS 2019 held in Qatar. She has co-authored articles in both database and ML communities and served as a member of the Program Committee for several international database conferences. She has recently received the best demonstration award at ICDE 2022 for her work on training data generation for learning-based query optimization.

Reproducible Query Optimization Research for Data Systems

Identifying reasonably good plans to execute complex queries in large data systems is a crucial ingredient for a robust data management platform. The traditional cost-based query optimizer approach enumerates different execution plans for each individual query, assesses each plan based on its costs, and selects the plan that promises the lowest execution costs. However, as we all know, the optimal execution plan is not always selected, opportunities are missed, and complex analytical queries might not even work. Thus, query optimization for data systems is a highly active research area, with novel concepts being introduced continuously. A wide range of proposals, from novel cardinality estimation methods to alternative physical operation selection strategies, have been proposed. However, qualitatively and quantitatively assessing their individual strengths and weaknesses is almost impossible. We thus introduce PostBOUND, a novel optimizer development and benchmarking framework that enables rapid prototyping and common-ground comparisons, serving as a community base for reproducible optimizer research.

Wolfgang Lehner is full professor at TU Dresden, Germany, leading both the Database Technology Group and the Institute of System Architecture. He is primarily interested in cross-cutting data management themes, ranging from complex analytical tasks and workflows to technologies that push the envelope in compiling and runtime of a data system. He is serving the international database community in various capacities (e.g., Vice-President of the VLDB Endowment, PVLDB Management Editor, and PC Co-Chair/MetaReviewer/Reviewer activities). He is an appointed member of the German Council for the Sciences and Humanities as well as a member of the Academy of Europe. He also holds a part-time professorship at the University of Aalborg, Denmark, and is a Fellow of the ACM.

Keynote Title: To Be Announced

Maria-Esther Vidal is a Professor at Leibniz University Hannover and Head of the Scientific Data Management (SDM) group at TIB–Leibniz Information Centre for Science and Technology. She is also affiliated with the L3S Research Centre and holds the title of Full Professor (retired) at Universidad Simón Bolívar (USB), Venezuela. Her research focuses on semantic data management, data integration, federated query processing, and machine learning over knowledge graphs, with a strong emphasis on neuro-symbolic AI. She has pioneered methods that integrate symbolic reasoning with machine learning to achieve interpretable and trustworthy AI systems. This work has had a profound impact in domains such as medicine (e.g., supporting diagnosis and personalized treatment), bias detection and documentation (ensuring fairness and transparency in AI pipelines), and scientific data ecosystems (enabling interoperability and reproducibility across heterogeneous data sources).