Uploaded June 2025 | Updated September 2026, 1 week ago
The National Research Platform (NRP) addresses key challenges in modern cyberinfrastructure, including architectural fragmentation in the post-Moore’s Law era and increasing demand for research resources. At its core is Nautilus, a globally distributed Kubernetes cluster spanning three continents, equipped with over 1,400 GPUs, more than 50 FPGAs, and diverse storage systems (Ceph, S3, OSDF). Its open architecture and tools, including JupyterHub and Coder, enable seamless collaboration across scientific disciplines and geographic boundaries. This presentation introduces the NRP and its provisioning model, with a focus on its network infrastructure and role as a research testbed. We highlight capabilities for smart network experimentation, including AutoGOLE/SENSE integration to enable automated, high-speed provisioning across various backbones and support interoperability with multiple federated testbeds. We also showcase our demonstrated reproducible use of P4-programmable FPGA SmartNICs for edge programmability, segment routing, and high-precision per-packet telemetry.
Mohammad Firas Sada: Mohammad Firas Sada is a researcher at the San Diego Supercomputer Center (SDSC) and the University of California, San Diego, specializing in high-performance computing (HPC), deep learning, and advanced networking. He leads initiatives within the National Research Platform (NRP), focusing on integrating FPGAs and SmartNICs into scalable AI workflows, enhancing network programmability, and developing FPGA High-Level Synthesis (HLS) solutions for HPC applications. Sada also spearheads the Seam Project, enabling federated HPC workflows across platforms like FABRIC, Chameleon Cloud, and CloudLab. His collaborative efforts with industry partners, such as Qualcomm, aim to deploy domain-specific transformer models on specialized hardware like the Cloud AI 100 Ultra cards. Committed to education and outreach, he mentors interns through programs like Google's Summer of Code and contributes to workshops and conferences, including FABRIC KNIT and PEARC, sharing insights on networked science and programmable networking.
Speaker Mohammad Firas Sada - San Diego Supercomputer Center
Dmitry Mishin
The National Research Platform (NRP) addresses key challenges in modern cyberinfrastructure, including architectural fragmentation in the post-Moore’s Law era and increasing demand for research resources. At its core is Nautilus, a globally distributed Kubernetes cluster spanning three continents, equipped with over 1,400 GPUs, more than 50 FPGAs, and diverse storage systems (Ceph, S3, OSDF). Its open architecture and tools, including JupyterHub and Coder, enable seamless collaboration across scientific disciplines and geographic boundaries. This presentation introduces the NRP and its provisioning model, with a focus on its network infrastructure and role as a research testbed. We highlight capabilities for smart network experimentation, including AutoGOLE/SENSE integration to enable automated, high-speed provisioning across various backbones and support interoperability with multiple federated testbeds. We also showcase our demonstrated reproducible use of P4-programmable FPGA SmartNICs for edge programmability, segment routing, and high-precision per-packet telemetry.
Mohammad Firas Sada: Mohammad Firas Sada is a researcher at the San Diego Supercomputer Center (SDSC) and the University of California, San Diego, specializing in high-performance computing (HPC), deep learning, and advanced networking. He leads initiatives within the National Research Platform (NRP), focusing on integrating FPGAs and SmartNICs into scalable AI workflows, enhancing network programmability, and developing FPGA High-Level Synthesis (HLS) solutions for HPC applications. Sada also spearheads the Seam Project, enabling federated HPC workflows across platforms like FABRIC, Chameleon Cloud, and CloudLab. His collaborative efforts with industry partners, such as Qualcomm, aim to deploy domain-specific transformer models on specialized hardware like the Cloud AI 100 Ultra cards. Committed to education and outreach, he mentors interns through programs like Google's Summer of Code and contributes to workshops and conferences, including FABRIC KNIT and PEARC, sharing insights on networked science and programmable networking.
Speaker Mohammad Firas Sada - San Diego Supercomputer Center
Dmitry Mishin
