Related Experiment Video
Updated: Jan 7, 2026
08:16
High-resolution Structural Magnetic Resonance Imaging of the Human Subcortex In Vivo and Postmortem
Published on: December 30, 2015
15.8K
Hierarchical Storage Management in User Space for Neuroimaging Applications
Valérie Hayot-Sasson1, Tristan Glatard2
1Department of Computer Science and Software Engineering, Concordia University, Montréal, Québec, Canada.
Neuroinformatics
|December 23, 2025
Summary
Sea, a new library, optimizes neuroimaging data transfers by intercepting read/write calls. It significantly speeds up data-intensive processing, especially when shared file systems are slow, without impacting performance when systems are not overloaded.
Area of Science:
- Neuroscience
- Computer Science
- Data Science
Background:
- Open-data initiatives increase neuroimaging dataset availability.
- Processing large datasets shifts bottlenecks to data transfer, challenging existing tools.
- Rewriting applications is a barrier to adapting neuroimaging tools for data-intensive processing.
Purpose of the Study:
- To develop a library, Sea, for efficient data management in standardized neuroimaging tools.
- To minimize data transfer time for large neuroimaging datasets.
- To evaluate Sea's performance on high-performance computing clusters.
Main Methods:
- Developed Sea, a library to intercept and redirect application read/write calls.
- Tested Sea on three neuroimaging preprocessing pipelines.
- Evaluated performance on three distinct neuroimaging datasets across two HPC clusters.
Main Results:
- Sea achieved speedups up to 32× on deteriorated shared file systems.
- Performance remained unaffected when shared file systems were not overburdened.
- Sea demonstrated minimal overhead in optimal conditions.
Conclusions:
- Sea effectively mitigates data transfer costs in data-intensive neuroimaging analysis.
- The library facilitates adaptation of existing tools without complete rewrites.
- Sea offers significant performance gains under specific, common, high-load conditions.
