Keynote Speakers
|
|
|
| Gilad Shainer | Omar Baldonado | Bilal Riaz |
| Senior Vice President Marketing at NVIDIA | Senior Director of Networking at Meta | Senior Director of Product Line Management and Head of Interconnect Strategy, Ciena |
|
Networking Innovations for Gigascale AI Systems [details] |
Lessons from networking Meta's gigawatt-scale AI fleet [details] |
The Future of AI Interconnects: Open Approaches to High-Performance AI Infrastructure [details] |
Panel
|
Moderator: Timothy Crawford, Managing Director at KeyBanc Capital Markets Abstract: Inference is moving in a big way from the Core to the Edge, in the form of Edge data centers (and smaller things at the edge), whether operated by Neoclouds, specialized operators, industries, or enterprises. How do the interconnects within and between these Edge data centers differ from those of the Core? Where do the constraints of power and latency influence the architectures and technologies for Scale-Up, Scale-Out, and Scale-Across? When do they operate cooperatively versus standalone? When is the connection to the Core considered Scale-Across or just the WAN? In this panel we will examine this critical, emerging venue for interconnects and learn what it can tell us about these classifications and applicable technologies. [details] |
Moderator
Timothy Crawford
Managing Director, KeyBanc Capital Markets
Inder Monga
Director of Berkeley Lab’s Scientific Networking Division and Executive Director of Energy Sciences Network (ESnet)
Yosuke Aragane
Vice President, IOWN Development Office, NTT
Tanner Ryan
Network Engineer at Cloudflare
Marc Austin
CEO and Founder - Hedgehog
Ofer Shapiro
Investor
Lightning Session
The week before OFC (the Optical Fiber Communication Conference) held in Los
Angeles in March, four new Multi-Source Agreements (MSAs) were announced.
Each is designed to accelerate design choices for some aspect of the AI data
center, ideally in a way that serves the needs of the operator customers.
They are not intended to give any vendor an advantage over the others but are
meant to speed deployment by getting the suppliers (sources) to agree on some
not-very-controversial aspects of these solutions. For each MSA, we have
asked one of the founding companies to speak and to explain what problem the
MSA is solving, why all the existing standards bodies and consortia are not
solving that problem, what the MSA promises to actually produce, how the MSA
will avoid stifling innovation, and how companies can join the MSA.
Drew Alduino
Meta
OCI MSA
Optical compute interface
Sponsor Talks
Alex Omø Agerholm
Chief Product Architect, Napatech
AI Fabrics in Motion: Programmable Interconnects for Scale-Up, Scale-Out, and Scale-Across
Nick Harris
Founder and CEO of Lightmatter, Lightmatter
BiDi DWDM: The Interconnect Foundation for High-Performance AI Training and Inference
Ravi Mahatme
Senior Director, Product Management, Photonic Fabric Business Unit, Marvell
A Shared Memory Architecture for Token-Efficient AI Infrastructure
Mohan Kalkunte
Vice President of Architecture & Technology, Core Switching Group, Broadcom
Distributed Computing @ Scale for AI Training & Inference
Will Eatherton
SVP, Data Center, Internet & Cloud Infrastructure Engineering at Cisco
Ethernet for Shared AI Fabrics: InfiniBand-Class Performance with Multi-Tenant Scale
Rohan Gandhi
Vice President of Optical Marketing at Eliyan
From Long Reach to Smart Reach: The Evolution of SerDes for AI Infrastructure
