Reproduce “IOSage: Detecting HPC I/O Bottleneck Patterns to Guide LLM Recommendations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mahdi Banisharif, Tabassum Mahmud, Ahmad Tarraf, Suren Byna, Ali Jannesari
IOSage helps HPC teams turn Darshan logs into application-specific I/O recommendations. A classifier selects among seven registered behaviors, then an LLM combines those patterns with benchmark-derived guidance, per-file evidence, and settings. Claude Sonnet 4 named a helpful change in 5 of 6 application scenarios; classifier Micro-F1 was 0.929 on 488 benchmark runs, falling to 0.832 on 158 outside-training runs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.