The move was announced at the Open Compute Project’s annual conference. Meta, formerly known as Facebook, outlined the following on its open networking gear advances. Wedge 400/400C top-of-rack switches. Meta collaborated with Broadcom and Cisco to deploy two top-of-rack switches called Wedge 400 and 400C. The Wedge 400 uses Broadcom’s Tomahawk 3 ASIC and 400C uses Cisco’s Silicon One. Celestica makes the networking equipment for Meta. Key points:
Both switches have more front panel port density for AI and machine learning applications.The systems have 12.8 Tbps switching capacity, four times the capacity of the Wedge 100S. Both switches have a field replaceable CPU system.
Meta’s FBOSS network operating system for its network switches now is using OCP’s SAI at scale. The move will allow Meta to work with more chip vendors. Also see:
Samsung begins mass production of OCP compliant enterprise SSDsDatacenter tech: Here’s where the Open Compute Project is going nextTarget joins Open Compute Project, aims to expand edge computing use cases
200Gbps/400Gbps data center fabrics. Meta has deployed 200G optics in its data centers and plans to deploy 400G in the future. Meta developed two next-gen 200G fabric switches with the Minipack2, a modular network switch, and the Arista 7388X5 with Arista Networks. Overall, Meta is making the argument that its open compute data center improvements will advance its plans to build out AI and its metaverse in a disaggregated model. More:
Facebook CEO Zuckerberg renames company Meta, outlines vision for ‘Horizon’ metaverseFacebook announces $50 million investment in ‘responsible’ metaverse developmentFacebook: Here comes the AI of the MetaverseFacebook brings metaverse to work with Horizon Workrooms (and you thought Zoom fatigue was bad)Why I will never use Zuckerberg’s metaverseWhen the metaverse comes, there are few good options on who will control it
