SC26: Renci to Showcase Advances in Research Networking, Data Management, and Scientific Workflows
With iRODS updates
This is a Press Release edited by StorageNewsletter.com on November 19, 2026 at 2:00 pmThe Renaissance Computing Institute (Renci) at the University of North Carolina at Chapel Hill shared advances in research networking, open-source data management, and AI-assisted scientific workflows at SC26. Conference attendees can visit Renci booth #1655 from November 17-19 at McCormick Place to meet researchers and collaborators and see how they are helping scientific teams move, manage, and analyze data across institutions.
iRODS updates for research data management
Terrell Russell and Kory Draughn presented three updates from the Integrated Rule-Oriented Data System (iRODS), an open-source, policy-based data management platform developed at Renci and supported by the iRODS Consortium.
- iRODS 5.1.0. The team highlighted logical storage quotas, FIPS compatibility, TLS support during setup, a new authentication scheme, and security and performance improvements. Talks are scheduled for Tuesday at 10am and Thursday at 11:30am CT
- iRODS System Tray App. A new desktop app for Windows and macOS automates the transfer of files from lab instruments into research data workflows. Talks are scheduled for Tuesday at 1pm and Wednesday at 2pm CT
- iRODS S3 API. The updated API gives teams more ways to access iRODS-managed data with S3-compatible tools, including support for public buckets and authentication using key pairs or tokens. Talks are scheduled for Wednesday at 3pm and Thursday at 11am CT
Fabric capabilities and research from its users
Fabric Principal Investigator Paul Ruth led booth sessions Tuesday from 3 to 4:30pm and Wednesday from 11:15am to 12:30pm CT. He discussed new capabilities for the NSF-funded programmable research infrastructure, including LoomAI, an interface that turns a plain-language description into an experiment that can be deployed, managed, and packaged across distributed resources. LoomAI already runs on Fabric, Chameleon, and Sphere. Ruth also discussed plans for a site in Cape Town that would extend Fabric into Africa.
Both sessions featured Jeff Weekley of UC Santa Cruz, who described how his team moves cryo-electron microscopy data to South Africa and processes it in transit, allowing researchers to assess results and adjust instrument settings before the full transfer is complete. Tuesday’s session also featured Anita Nikolich of the University of Illinois on post-quantum cryptography. On Wednesday, Nik Sultana of Illinois Tech shared a Fabric user story.
Visitors also get the chance to learn more about the development of FAB-Cache, a programmable service built on Fabric that helps research teams retrieve datasets from remote sites, originally presented at Sunday’s Innovating the Network for Data-Intensive Science (INDIS) workshop.
Renci researchers co-author three papers at Works 2026
Anirban Mandal, director of Renci’s Network Research and Infrastructure Group, co-organized the 21st Workshop on Workflows in Support of Large-Scale Science (Works 2026) on Sunday, Nov. 15. The program featured three papers co-authored by Renci researchers, including a Distinguished Paper on when to use large language models in high-performance computing schedulers.
- Decentralized Scheduling of Scientific Workflows: Renci co-authors Komal Thareja and Mandal worked with collaborators to develop a scheduler that lets distributed agents agree on task assignments without a central coordinator. In a test on the MareNostrum 5 HPC, the approach shortened workflow execution time and distributed work more evenly across agents
- RLSA: Adaptive LLM Gating for Multi-Agent HPC Scheduling Agents. Co-authored by Mandal, this Distinguished Paper examined when scheduling agents can use a faster alternative to a large language model. The approach nearly halved model calls per job in a controlled experiment. The team also evaluated it on Fabric
- Complexity and Scale in AI-Assisted Workflow Management: A Federated Learning Case Study. Thareja, Mandal, and collaborators evaluated whether AI agents could build and validate a medical imaging workflow spanning more than 2,600 jobs across four sites. In experiments on Fabric, checking the generated workflow vs. explicit requirements caught errors that debugging based on runtime failures had missed
More from Renci and its collaborators
The booth program also featured work on Swarm, PegasusAI, and CI Compass, three projects involving Renci’s Network Research and Infrastructure Group, as well as MESA, an open-source platform that uses AI agents and curated metadata to help researchers find and work with scientific data. On Wednesday at 4pm CT, Renci’s Stephanie Suber discussed how the Cyberinfrastructure Community-wide Mentorship Network (CCMNet) connects professionals across institutions for mentorship and practical guidance.
Visit Renci at booth #1655 to attend a talk, meet the teams, or discuss a potential research collaboration.













