Full text
Modernizing the DesignSafe Researcher Dashboard Portal Vani Walvekar∗, Sameera Sineen∗ ∗Saint Louis University Email: v[email protected], [email protected] Abstract As part of the SGX3 Internship Program at the Texas Advanced Computing Center (TACC), our team collaborated with the DesignSafe group to reimagine and enhance their researcher dashboard. DesignSafe is a widely used science gateway supporting natural hazards research by providing access to HPC systems, data, and tools. The dashboard serves as the main entry point for researchers to manage jobs, allocations, and storage. Our work focused on modernizing the dashboard interface to make it more intuitive, informative, and helpful. Beyond design, we also introduced two intelligent backend features: one to predict the likelihood of job failure before submission, and another to assist users with cleaning and managing their data using machine learning. These improvements aim to reduce researcher frustration, increase productivity, and create a smoother interaction with the gateway. Index Terms High-Performance Computing (HPC), Science Gateway, Dashboard Design, Machine Learning, Job Failure Prediction, Data Cleanup, User Experience, Natural Hazards Research, DesignSafe, Researcher Tools Key Features of the DesignSafe Modernized Portal: •Redesigned Dashboard Experience: A clean, modular layout shows recent jobs, allocations, favorite tools, and system status in one view. Unlike the older version, which required users to navigate through multiple sections, the updated interface provides an at-a-glance overview for managing work efficiently. •Job Failure Prediction Using Machine Learning: A prototype analyzes job parameters using models like XGBoost and CatBoost to estimate the likelihood of job failure before submission. This feature parses historical logs and provides a real-time risk score, saving time and HPC resources. •Smart Storage Analyzer for Data Cleanup: Integrated with the Data Depot, this tool uses Isolation Forest to detect duplicate or outdated files and suggest cleanup actions via a user-friendly UI. •Personalized Experience with Favorite Apps and Guides: Users can pin frequently used tools and access embedded tutorials directly from the dashboard, making the platform more personalized and accessible. •Real-Time System Status and Tool Access: Leveraging the Tapis API, researchers can view current system loads and access recent tools quickly-helping them optimize job submission timing. In conclusion, the redesigned DesignSafe portal brings modern UI improvements and intelligent backend tools to support researchers in high-performance computing environments. With features like real-time system visibility, job failure prediction, and AI-assisted storage management, the portal enhances usability and scientific productivity. This project highlights how thoughtful design and applied machine learning can transform researcher interaction within large-scale scientific platforms.
