Today (22 January) marks the final day of this systems-focused conference, which emphasizes a systems architecture perspective and attracts practitioners from both academia and industry. Among the authors presenting their best systems research were Peter Boncz (leader of the DA group and general co-chair of CIDR) and Hannes Mühleisen (CWI & DuckDB Labs)*.
Data systems take centre stage at CIDR 2025
The Conference on Innovative Data Systems Research (CIDR) brought 170 international visitors to Amsterdam this year. Organized locally by our Database Architectures (DA) group, CIDR is renowned for showcasing bold ideas, novel approaches, and forward-thinking perspectives in data systems research.
Publication date
22 Jan 2025
Share this page
Working with vast amounts of data
Keynote speaker Thea Klaeboe Aarrestad (ETH) discussed the immense challenges of managing the vast amounts of data produced by the CERN Large Hadron Collider, where researchers study high-energy proton collisions to uncover fundamental particles and forces in physics. These collisions generate 10,000 exabytes of raw data annually. To handle this enormous data flow, real-time event filtering systems process millions of collisions per second, equivalent to 5% of global internet traffic, combined with unprecedented complexity.
Maintaining data integrity for meaningful physics analyses requires highly optimized machine learning (ML) algorithms. Klaeboe Aarrestad explained how real-time ML techniques enhance data filtering and improve the acceptance of physics signals.
Energy trade-offs
Keynote speaker Christos Kozyrakis (Stanford University) explored the limits of semiconductor technology and emerging trends in AI training and inference that are reshaping hardware architecture. He highlighted the energy trade-offs in compute and memory, offering insights into how AI systems and chips can also accelerate other data-intensive workloads and influence the development of future hardware technology.
Pushing boundaries
In addition to keynotes, the program included an evening programme, paper and demo presentations. CIDR values bold and innovative approaches to data management system architecture, real-world systems-building experience, insightful experimental studies, and provocative position papers. The conference fosters discussions around visionary ideas that push the boundaries of what is possible in data systems.
About CIDR
CIDR began in 2003 as a biennial conference and quickly became the premier event for practical data systems research. Since 2020, it has been held annually, alternating between Amsterdam and the USA. This year the conference was held from 19 to 22 January.
*Adaptive Factorization Using Linear-Chained Hash Tables
Paul Gross, Daniël ten Wolde, Peter Boncz
*Runtime-Extensible Parsers
Hannes Mühleisen, Mark Raasveldt (DuckDB Labs)
Pictures: copyright CIDR
