Speaker
Description
ROOT RNTuple I/O subsystem has been designed to address performance bottlenecks and shortcomings of ROOT's current state of the art TTree I/O subsystem. RNTuple provides a backwards-incompatible redesign of the TTree binary format and API that evolves the ROOT event data I/O for the challenges of the upcoming decades. It has been engineered for high-performance on modern storage hardware, a compact data format, and features a robust and easy to use interface. RNTuple currently provides many tunable parameters (e.g. page size, target compressed cluster size, number of readahead clusters scheduled) that can be adjusted for specific use cases.
This contribution explores the RNTuple parameter phase space using machine learning
techniques, trying to gain insights as to whether the ideal value for these parameters is analysis-dependent or otherwise general. This question is important for performance engineering, as many parameters drastically impact read/write performance and memory footprint.
| Consider for long presentation | No |
|---|
