Published January 20, 2026
The FABRIC’s integration with Chameleon GPUs is now available, giving researchers a powerful new way to run distributed experiments using FABRIC’s programmable, high-performance networking with Chameleon’s powerful GPUs.
What’s Changed
Previously, Chameleon users could select FABRIC as a backend network without logging into FABRIC. While convenient, this limited how much control researchers had over their networking.
Now, researchers explicitly use FABRIC to:
- Design custom network topologies
- Access advanced programmable networking hardware
- Efficiently move large datasets across sites
- Set up workflows linking distributed data resources with Chameleon GPUs
This makes the network an active part of the experiment, not just connectivity.
Ready to Use Today
The integration is fully deployed on Chameleon and available in production. Researchers can:
- Create Chameleon and FABRIC accounts
- Select and deploy the integration environment through the FABRIC Artifact Manager.
- Use FABRIC as the network fabric across multiple Chameleon sites
A full walkthrough is available via a webinar led by Ben Formby, linked here.
Why It Matters
This integration enables:
- High-performance data movement
- Cross-site experiments
- Programmable network research
- Infrastructure-aware workflows
Researchers can now experiment not just with compute, but with the network itself.