Biography
Wozniak designs and implements workflow systems for scientific applications and deep learning workloads, combining techniques from high-performance computing and distributed computing. He has extensive experience integrating advanced computing techniques in collaboration with the health, experimental, and simulation sciences.
SRP Project Title
Applying Performance Prediction to Live Filesystems
Topical Areas
Applied Computer Science; Applied Mathematics; Artificial Intelligence and Intelligent Systems; Open Source Software; Performance Evaluation and Benchmarking; Visualization and Human-Computer Systems
Abstract
Large shared filesystems can be an unpredictable environment for running large scientific workflows. Our team has developed a performance collection and prediction software package. We would now like to improve the software, make it more portable and easy to use, and run it on more real-world filesystems. The team in this project will run the data collection and prediction tools on live systems with real applications. The research aspect will be in how to interpret and visualize the resulting data. Possible applications include light source experiments from the Advanced Photon Source and data management workflows from epidemics modeling.
Desired Skills
Linux Shell scripting Machine learning
Lightning Talk Title
Prediction and resilience for distributed workflows
Keywords
workflows;hpc;filesystems;prediction