How Fuzzball Unifies HPC and AI Workloads on One Infrastructure

David Godlove shows how Fuzzball runs traditional HPC jobs, AI and ML workloads and interactive sessions on one unified infrastructure. The demo starts in the ready-to-run workflow catalog, which holds templates for HPC applications such as GROMACS and LAMMPS, newer AI workflows such as NVIDIA inference microservices and AlphaFold 2, and interactive tools such as Jupyter notebooks and ParaView.

Selecting a template, accepting or adjusting its default parameters and clicking run is all it takes to start a workflow. Interactive applications can be launched and connected to from the Fuzzball web interface, and Fuzzball places each job on appropriate resources based on its computational requirements and data locality.

Key takeaways

• The Fuzzball workflow catalog includes ready-to-run templates for GROMACS, LAMMPS, NVIDIA inference microservices, AlphaFold 2, Jupyter and ParaView.
• Each template ships with default parameters that can be accepted as-is or modified before clicking run.
• CPU-based HPC jobs run alongside AI and ML applications, and interactive apps can be connected to from the web interface.
• Fuzzball places jobs on resources automatically based on computational requirements and data locality.

Questions this video answers

Can Fuzzball run HPC and AI workloads on the same infrastructure?

Yes. The demo launches traditional CPU-based HPC applications like GROMACS and LAMMPS, AI and ML workflows such as NVIDIA inference microservices and AlphaFold 2, and interactive Jupyter notebooks and ParaView sessions from one catalog. Fuzzball places each job on suitable resources according to its computational needs and data locality.

About this video

What does it actually look like to run traditional HPC jobs, AI training workloads, and interactive notebooks on the same platform? In this demo, CIQ's David Godlove shows exactly how Fuzzball handles it. Starting from Fuzzball's ready-to-run workflow catalog, David walks through:

Selecting and launching pre-packaged workflow templates for traditional HPC applications like GROMACS and LAMMPS Running modern AI and ML workloads including NVIDIA Inference Microservices and AlphaFold 2 Launching interactive environments like Jupyter notebooks and ParaView directly from the Fuzzball web UI How Fuzzball intelligently places jobs on the right resources based on computational requirements and data locality — no manual intervention required

Whether you're a sysadmin looking to consolidate HPC and AI infrastructure or an AI/ML engineer who needs flexible, policy-driven access to GPU and CPU resources, Fuzzball gives you a single platform with the orchestration intelligence to handle it all.

This video is part of the Fuzzball playlist. Browse every CIQ video by product and topic.