What is it about?
Cloud computing makes it possible to run large-scale data analysis using flexible computing resources. But users often struggle to decide how much computing power to allocate. More resources usually mean better performance, but also higher costs. Today’s cloud databases require users to fix resource settings before running a query. If performance or cost is unsatisfactory, the only option is to wait, or kill the query and try again, wasting both time and money. Our research introduces Intra-Query Runtime Elasticity (IQRE), a new capability that allows users to dynamically adjust computing resources while a query is running, without interruption. We built Accordion, the first system to realize IQRE in a cloud-native SQL engine. It provides intuitive controls and automatic tuning to help users meet performance goals at minimal cost. IQRE lays the foundation for a new generation of serverless, intelligent cloud databases—systems that adapt automatically to changing workloads, optimize resource usage in real time, and make large-scale data analysis faster, smarter, and easier for everyone.
Featured Image
Photo by Markus Winkler on Unsplash
Why is it important?
Cloud databases give users access to massive computing power, but they still require performance and resource decisions to be made before a query starts. Once a query is running, it’s usually too late to adjust how fast it runs or how much it costs—unless the user cancels it and starts over. This work introduces Intra-Query Runtime Elasticity (IQRE), a new capability that allows users to change the performance and resource usage of a query while it is still running, without interruption. Our system, Accordion, is the first engine to support this behavior in practice. IQRE opens the door to a new generation of cloud-native databases that are more flexible, efficient, and intelligent—systems that respond in real time to user needs, reduce waste, and simplify the process of analyzing data in the cloud.
Perspectives
Working on this paper has been a rewarding journey. I’ve long been interested in how database systems can become not just faster, but also more intelligent—more responsive to users, and more aligned with real-world needs. This work started from a simple idea: why should query tuning stop once execution begins? With Intra-Query Runtime Elasticity (IQRE), I wanted to challenge that assumption and give users more control—to make data analysis more flexible, more intuitive, and ultimately more free. I'm especially proud that Accordion was built from scratch, and that it brings this concept to life in a practical and usable system. I hope this work will inspire others to rethink what query execution can be, and how cloud databases can better serve human goals, not the other way around.
XUKANG ZHANG
Renmin University of China
Read the Original
This page is a summary of: Intra-Query Runtime Elasticity for Cloud-Native Data Analysis, Proceedings of the ACM on Management of Data, June 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3725315.
You can read the full text:
Contributors
The following have contributed to this page
