Click Here to View the Recorded Webinar
(Registration is still required to view this FREE Webinar)
 |
 |
Duration: 1 Hour
Description:
The “network wall” has become one of the primary bottlenecks in large-scale AI training and inference. While compute FLOPS have doubled every 18 months, merchant Ethernet switch ASIC capacity and radix (the number of compute devices that can be connected) has historically doubled in steps longer than two years. State-of-the-art switches support ~100 Tbps today and are only expected to reach ~200 Tbps by 2028. Switch radix in turn limits the single-hop domain size, which either limits the scale or extends the training time of sparse Mixture-of-Experts (MoE) models. A new switch technology and network architecture is required to deliver an order-of-magnitude increase in radix, breaking the network wall, and enabling AI to scale more effectively.
Key topics covered:
- Scaling challenges in AI data centers: compute, memory, networking and power
- The network wall: why switch radix has become a critical constraint
- The high-radix switching opportunity: building flatter networks for larger clusters and higher performance
- Impacts of high-radix switching on AI training + inference
What will you learn?
- AI Networking market overview and potential for high radix switching
- Why switch radix has become a critical constraint in scaling AI infrastructure
- How to use high-radix switching to build flatter, more scalable networks and larger clusters
- How to achieve higher AI training + inference scale and performance using high-radix networks
Who should attend?
AI infrastructure architects, engineers and operators, including hyperscalers, neoclouds, sovereign AI entities, and large enterprises who are building AI training and/or inference data centers.
Speakers:
Cliff Grossner, Ph.D.
Chief Innovation Officer, Open Compute Project Foundation (OCP)
At the Open Compute Project Foundation, Dr. Grossner leads its market intelligence function and is responsible for driving awareness of OCP, training and certification programs, and guiding inventors presenting their early-stage company ideas to potential investors. Dr. Grossner has more than 25 years of telecommunications industry experience encompassing scientific research, market analysis, corporate and product strategy, product management and marketing. His current interests include IT Ecosystems from cloud to the edge looking at the equipment [servers, IT networking and storage] and software, and physical infrastructure [power and cooling] needed for computation, with a focus on edge computing, silicon components included in IT equipment including coprocessors for AI and ML. Previously, Dr. Grossner was head of the Cloud and Data Center Research Practice, which he launched at Infonetics research and followed through transitions to IHS, IHS Markit and Informa Group. Prior to Infonetics, he held senior positions, including heading strategic marketing for Alcatel-Lucent’s enterprise network business, tenures at Bell Labs, several startups and Nortel. He earned his Ph.D. at McGill University, and his Master of Science in Computer Science at Concordia University, winning national scholarships to support his graduate work. He holds more than 10 patents in computer networking, networking embedded security and telecommunications applications.
Alan Weckel
Founder and Technology Analyst, 650 Group
Alan is a recognized data center networking, enterprise, and Cloud market expert. He publishes quarterly market share and forecast data for switching, routing, and Cloud markets. His research spans verticals, use case, and speed transitions. In addition to specific technology reports, Alan publishes on adjacent markets such as AI Networking, DCI, and Data Center Semiconductors and is a market influencer in the financial community and standards bodies. Alan is at the forefront in market methodology and segmentation, pioneering research in areas such as white box, data center switching, AI, Cloud CAPEX, and merchant silicon.
Promode Nedungadi
Chief Technology Officer, Eridu
Promode Nedungadi is a technology executive and innovator whose deep expertise has helped to drive breakthrough architectures and translate complex technologies into scalable, real-world solutions and multiple billion-dollar platforms.
Across his career, Promode has driven and shaped platforms at some of the most influential companies in networking and infrastructure. His work has included helping to build one of the industry’s first enterprise network delivery as-a-service platforms at Nile Global, leading work on next-generation packet-optical and transport architectures at Infinera, driving cloud and distributed storage system strategy at Seagate, and working as Chief Architect responsible for high-end routing platforms including the ASR 9000 during his 15 years at Cisco.
A prolific inventor with more than 40 patents, Promode continues to influence next-generation networking at Eridu, where he leads silicon, software, and systems architecture and strategy.
Jay Gill
Sr. Director of Product Marketing, Eridu
Jay Gill is Sr. Director of Product Marketing at Eridu. Prior to joining Eridu, Jay held senior marketing roles at HPE Juniper Networks, Pluribus Networks (acquired by Arista) and Infinera, and strategy and product management roles at Cisco Systems. Jay previously worked in engineering and product development at multiple incumbent and startup service providers. Jay holds a BSEE and MSEE from Stanford and an MBA from UCLA Anderson.