Tuesday, October 17, 2023

#OCPSummit23: AI is Open Compute Project's Biggest Challenge

Is AI the biggest challenge for the Open Compute Project (OCP)? 

Zaid Khan, Board Chair from Open Compute Project, explains:

- The sudden pivot to AI caught everyone by surprise, leading to a race in optimizing systems, data centers, and infrastructure to accommodate this shift.

- The lack of collaboration among companies and vendors in addressing AI challenges presents a unique opportunity for OCP to bring everyone together and solve these problems as a community.

- OCP is embarking on an important journey to tackle these challenges, emphasizing that this is not a problem one company can solve alone, but requires collective effort and collaboration. 


Filmed at OCP Summit 2023 (#OCPSummit23) in San Jose. Want to be involved our video series? Contact info@nextgeninfra.io

Google opens Falcon for Hardware Transport

Google is contributing its Falcon hardware transport protocol to the Open Compute Project.

Falcon is a hardware-assisted transport layer designed for high-performance, low-latency connections in Ethernet data center networks. 

It leverages production-proven technologies including Carousel, Snap, Swift, PLB, and CSIG.

Falcon can support RDMA and NVM Express as well as other Upper layer protocols (ULPs).

Some industry perspectives shared by Google:

“We welcome Google’s contribution of Falcon as it shares the Ultra Ethernet Consortium’s vision to drive Ethernet as the best data center fabric for AI and HPC, and look forward to continuing industry innovations in this important space.” - Dr. J Metz, Chair, Ultra Ethernet Consortium (led by AMD, Arista, Broadcom, Cisco, Eviden, Hewlett Packard Enterprise, Intel, Meta, Microsoft, and Oracle).

“Falcon is first available in the Intel IPU E2000 series of products. The value of these IPUs is further enhanced as the first instance of an Ethernet transport to add low tail latency and congestion handling at scale. Intel is a Steering Member of Ultra Ethernet Consortium, which is working to evolve Ethernet for high performance AI and HPC workloads. We plan to deploy the resulting standards-based enhancements in future IPU and Ethernet products.” - Sachin Katti, SVP & GM, Network and Edge Group, Intel.

"We are pleased to see a high-performance transport protocol for critical workloads such as AI and HPC that works over standard Ethernet/IP networks and enables massive application bandwidth at scale." - Hugh Holbrook, Group VP, SW Eng., Arista Networks.

“Cisco is pleased to see the contribution of Falcon to the OCP. Cisco has long supported open standards and believes in broad ecosystems. The rate and scale of modern data center networks and particularly AI/ML networks is unprecedented, presenting a challenge and opportunity to the industry. Falcon addresses many of the challenges of these networks, enabling efficient network utilization.” - Ofer Iny, Cisco Fellow, Cisco.

BT unveils its next-gen Global Fabric

BT unveiled its new Global Fabric network for connecting enterprise customers with multiple clouds.

BT says Global Fabric represents a generational shift in technology, based on a network-as-a-service (NaaS) technical and commercial model. 

The new network “underlay” will be an AI-powered digital orchestration layer. Global Fabric will be “deterministic”, selecting the optimal end-to-end paths for applications and workloads to move to and between multiple clouds and end users. Options will include a new BT-enhanced Internet service, point-to-point Ethernet, multi-point Ethernet and MPLS and these will be offered in bandwidth increments of 1 Mbps up to 100 Gbps. Connectivity will be interchangeable on the same port, offering flexibility not possible on current networks.

The fully programmable network is built with state-of-the-art equipment. BT estimates that when fully rolled out, Global Fabric will use 79 per cent less electricity than its current global networks.

Bas Burger, CEO, Business, BT, said: “Global Fabric will future proof customers’ connectivity by providing flexibility to ensure they’re always connected so they can always be productive. They’re facing a new wave of digital revolution with AI, IoT and automation driving demand for simplicity and better multi-cloud connectivity. Customers can achieve better total costs, boost app performance and user experience, all while complying with regulations and mitigating cyber threats. Global Fabric means multi-cloud works better on BT.”

Chris Sharp, Chief Technology Officer, Digital Realty, said: “Global Fabric is a great example of how service providers should be innovating and will offer enterprises new options for connecting to Digital Realty, the world’s largest data centre platform. By building a cloud-centric network and locating its PoPs inside our world-leading carrier neutral facilities (CNFs), BT will be able to offer its customers terrific speeds with low latency while helping them minimise their environmental impact.”

Brenden Rawle, Senior Director Business Development EMEA at Equinix said: "We welcome BT’s launch of Global Fabric, which builds on its Connected Cloud Edge solution launched with Equinix last year. We enjoy a great partnership with BT and look forward to working together to connect customers to the doorstep of the cloud with a vast and varied choice of potential partners and other SaaS providers, in the Equinix location of their choice."


Broadcom develops DOCSIS 4.0 AI Chipset with Comcast

Comcast and Broadcom are designing and building the new chipset based upon CableLabs’ DOCSIS 4.0 specifications and aimed at enabling the first AI-powered access network.

The new chipset that embeds AI and machine learning (ML) within the nodes, amps and modems that comprise the last few miles of Comcast’s network. Benefits are expected to include:

  • Making smarter network performance decisions using network diagnostics insights produced by both local and cloud AI.
  • Enhanced monitoring and issue detection using bandwidth-efficient telemetry data.
  • Transforming network maintenance of the network using real time issue localization plus predictive and self-healing network intelligence.
  • Protecting network facilities and customers with improved cybersecurity intrusion detection.
  • Assisting customers more effectively through local and cloud-based AI.
  • Monitoring home IoT devices for connectivity disruptions.

Additionally, the new chipset will be the first in the world to incorporate DOCSIS 4.0 Full Duplex (FDX), Extended Spectrum (ESD) and the ability to run both simultaneously, enabling Internet service providers across the globe to deliver DOCSIS 4.0 services using a toolkit with technology options to meet their business needs. 

“By enabling a toolkit that includes FDX, ESD or both simultaneously, this new Unified DOCSIS4 chipset incorporates the advantages of both technologies and will enable economies of scale as well as a common retail modem for the industry,” said Rich Nelson, Senior Vice President and General Manager, Broadband Video Group, Broadcom. “The edge network and in-home capabilities of this chipset will improve network intelligence and reliability to create an improved broadband experience with enhanced privacy protection and cyber security for the consumer.”

OCP and Ultra Ethernet Consortium team on AI Clusters

The Open Compute Project Foundation (OCP) and the Ultra Ethernet Consortium (UEC) announced a new collaboration to improve Ethernet performance for next-generation AI clusters and HPC. 

UEC plans to develop enhancements to Ethernet, and the OCP has an active Community developing sustainable large-scale computational infrastructure for AI and HPC with Ethernet. 

Key aspects of the collaboration will involve aligning work within OCP and UEC, focusing efforts on shared objectives and ensuring OCP's integration of UEC's Ethernet enhancements is smooth and effective. Overall, the alliance will leverage the expertise of both organizations to advance Ethernet performance for AI workloads. Areas that have been identified for initial exploration on potential collaborations include: OCP Switch Abstraction Interface (SAI), OCP Caliptra Workstream, OCP Networking Project, OCP NIC Workstream, OCP Time Appliance Project, and OCP Future Technologies Initiative.

“The uptake for AI to assist humans in the workplace is expected to be widespread, including automating data center operations, and has created a data center IT equipment investment perfect storm. Capex spend forecasts have been adjusted upwards by major analyst firms, driven by the planned deployments of hyperscale data center operators of large-scale AI clusters. By collaborating, the UEC and the OCP Community system specifications can have a greater influence on the development of sustainable, large-scale AI clusters addressing key memory size, and connectivity bandwidth challenges posed by large language models," said George Tchaparian, CEO at the Open Compute Project Foundation. 

“AI and HPC workloads present new network challenges such as the need for greater scale, more bandwidth, multi-pathing and faster reaction to congestion. UEC was formed to develop specifications for Ethernet that will bring improvements to better meet the requirements of these workloads as they are experiencing tremendous growth which will change the nature of network traffic, potentially as significant as the switch from voice to video. Collaborating with the OCP Community will fast-track the integration of the UEC-inspired Ethernet enhancements into complete systems that can serve the market”, said J Metz, Steering Committee Chair at the Ultra Ethernet Consortium.

MaxLinear launches DOCSIS 4.0 cable modem and gateway chipset

MaxLinear released its DOCSIS 4.0 SoC cable modem and gateway platform, Puma 8. 

The new chipset builds on the success of its DOCSIS 3.0 and 3.1 predecessors (Puma 6 and 7) and includes all critical elements of a DOCSIS platform, including a low-power, ultra-high split upstream PGA. The platform is compliant to the DOCSIS 4.0 frequency division duplex (FDD) specification and enables immediate large-scale deployments as a DOCSIS 3.1 device with gradual feature expansion options without the need for costly CPE replacements. 

Complete modem, Embedded Multimedia Terminal Adapter (EMTA), and gateway platforms will be available from MaxLinear’s early access OEM partners in 2024.

Several market leading OEM partners have developed MaxLinear-based DOCSIS 4.0 platforms in preparation of compliance and interoperability testing, field trials, and mass production deployments during 2024. MaxLinear’s early access partners include but are not limited to Askey, CommScope, and Hitron.

"We’ve harnessed our extensive IP portfolio and system knowledge to develop a comprehensive DOCSIS 4.0 system and platform, adopting a holistic strategy that encompasses a full suite of interrelated functionalities spanning every aspect of the system,” stated Will Torgerson, Vice President and General Manager of the Broadband Business Unit. “MaxLinear’s solutions empower service providers with innovation, flexibility, and reliable options.”