Daan de Graaf (TU/e) at GRADES-NDA ‘24
Our SciLake partner, Daan de Graaf, had the opportunity to present an accepted article on behalf of authors Wilco van Leeuwen, George Fletcher, and Nikolay Yakovets, all from Eindhoven University of Technology (TU/e). The team showcased "HomeRun", a tool specifically designed for comparing different cardinality estimation techniques in graph databases.
For those new to the topic, the cardinality of a graph database refers to the number of elements in a set, such as the number of edges connected to a node or the total number of nodes in the database. Accurate cardinality estimation is crucial for optimising the performance of queries, as it helps plan the most efficient way to retrieve data.
One of HomeRun's key features is its ability to evaluate the performance of different cardinality estimation techniques in given usage scenarios. The tool generates visualisations automatically, helping users understand the trade-offs between various techniques. This tool is particularly useful for database developers when they face performance issues, like long-running queries, with specific query and dataset combinations.
In SciLake, HomeRun is being used to optimise the database system performance in the context of the WP2 Data Lake Search and Navigation.
For more information about HomeRun, you can refer to the paper:
- Wilco van Leeuwen, George Fletcher, and Nikolay Yakovets. 2024. HomeRun: A Cardinality Estimation Advisor for Graph Databases. In Proceedings of the 7th Joint Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA) (GRADES-NDA '24). Association for Computing Machinery, New York, NY, USA, Article 6, 1–9. https://doi.org/10.1145/3661304.3661902