2025 Midwest RCD Annual Meeting

Quantum computing is poised to revolutionize the landscape of scientific research, offering unprecedented capabilities to solve complex problems beyond the reach of classical systems. In this talk, Dr. Daniel Blankenberg, a computational biologist at the Cleveland Clinic, will provide an accessible introduction to quantum computing, outlining its foundational principles and its growing role in healthcare and life science research.

A key focus will be the transformative partnership between Cleveland Clinic and IBM, highlighting how collaboration accelerates advancements. We discuss how quantum computing enhances research efforts across disciplines, from drug discovery to data analysis, sharing successes that showcase the potential for real-world impacts. Finally, the talk will explore the future of quantum computing including emerging challenges, promising developments, and the next steps for leveraging quantum technology to drive scientific breakthroughs. Whether you are new to quantum computing or an experienced researcher, this talk will offer valuable insights into how quantum innovation is shaping the future of life science research and beyond.

Several HPC centers are exploring how to provide and user LLM-based chat interfaces for both research and client support use cases. We will present, share lessons learned and answer your questions for three such projects in this session.

1. QA Tool for ACCESS - an AI-driven information retrieval system using Retrieval-Augmented Generation (RAG) and LLMs, delivering relevant, independent query responses while continuously improving through feedback and data updates.
2. AnvilGPT is a Generative AI platform hosted at Purdue University through the Anvil ACCESS resource. It offers researchers nationwide access to various open-source large language models via chat and API interfaces, aiming to enhance productivity and develop AI-enabled scientific workflows.
3. The hpcGPT project seeks to transform HPC user support by creating a generative AI-powered question-answering service. This conversational interface will address policy inquiries, technical suggestions, performance issues, and debugging.

PRESENTATION 1: Stephen Deems, Pittsburgh Supercomputing Center (12 mins)
Building On-Ramps to the NSF ACCESS Ecosystem The NSF ACCESS program connects your institution’s researchers and educators to a range of national-scale computing systems—at no cost to you or them. This talk will introduce campus IT staff to the ACCESS program and the services it offers. We will also describe ACCESS On-Ramps, a new feature developed by the ACCESS Allocations team that lets campus IT staff share ACCESS resource information with your campus community. Our goal is to help researchers find out about ACCESS in the first place they go for IT information—their campus IT websites.

PRESENTATION 2: Geoffrey Lentner, Purdue University (12 mins)
HyperShell: A user-facing tool for high-throughput scheduling of small jobs We’ve been crafting a user-facing tool for high-throughput computing scenarios for the past 5 years. The HyperShell software is a command-line tool that fits somewhere between GNU Parallel and HTCondor, with HPC user ergonomics and features in mind. It allows for users to run millions of small jobs, automate "pilot job" workflows, or make better use of large GPUs for packing smaller jobs in parallel. Currently, we are focused on user education, workshops, and outreach. The goal of this presentation will be to share the essential concepts and functionality of HyperShell with other center staff who may be in a position to facilitate researchers who might benefit from this capability. https://hypershell.org https://github.com/hypershell/hypershell

PRESENTATION 3: Joseph Tang, Ohio Super Computing Center (12 mins)
Backup Validation at the Ohio Supercomputer Center: A Belt and Suspenders Approach We present our current large data restoration and backup validation scheme at the Ohio Supercomputer Center (OSC).  There are approximate 12 PiB data and ~ 3 billion files stored in Project (ESS) and NetApp storage systems at OSC, and it is critical to efficiently restore data and validate the fidelity of backups. It is essential for maintaining business continuity, especially in disaster recovery scenarios.  Per NIST 800-53 compliance frameworks, we employ acceptance sampling concepts and statistical models estimating the amount of sample sizes, as it is not practical to restore 10+ PB of data with billions of files at once.  Projects / filesets are divided into collocation groups, so we could apply level of parallelism with group of nodes and reduce serialization due to large data volume from one node when migrating data from disks to tapes.  We employ “no query restore” from IBM Storage Protect with multiple sessions and restore data from filesets in the same collocation group, so the data could be restored efficiently.  Restored data are validated by comparing checksum of original files in snapshots with checksum of restored files using in-house developed scripts.  A typical restore test is 110 – 140 TiB data.  In the end, we discuss the current challenges and future directions about data restoration.

PRESENTATION 4: Preston Smith, Purdue University (12 mins)
Updates on statistical models for presenting the value of RCD investments
This presentation will discuss new research in modeling RCD value: specifically production function models across multiple institutions, and a work in progress effort for institutional benchmarking as a function of a university’s research output. Participants will learn best practices in carrying their value statement back to administration, and tools for benchmarking against peer institutions.


Q/A Session (~10 Minutes)

This session will be a mix of presentation and interactive discussion using Mentimeter.