MENU

World’s first 51.2 Tbit/s Ethernet switch chip plans chiplet optical version

Technology News |
By Nick Flaherty

Broadcom has shipped samples of the world’s first Ethernet switch capable of 50.2 Tbit/s, or 64 channels of 800 Gigabit Ethernet in data centres.

The 5nm BCM78900 monolithic chip is the first in the StrataXGS Tomahawk 5 switch series and has  double the bandwidth of any other switch silicon available for AI applications says the company.

This supports 64 × 800GbE, 128 × 400GbE, or 256 × 200GbE ports. The device family features a maximum of 64 integrated Peregrine SerDes cores, each with eight integrated 106-Gb/s PAM4 SerDes transceivers and associated PCS for native support of numerous physical connectivity options, enabling a broad range of media, speed and reach.

The BCM78900 architecture delivers complete L2 and L3 switching, routing, and tunneling capabilities at line rate and maximum port density, with low power and latency. The first version support direct copper connections without the need for retiming chips, and a chiplet version is planned with photonic interfaces in the same package.

Software compatibility is maintained across the StrataXGS product portfolio to simplify customer designs. The Tomahawk switches are already used by the world’s largest data centre operators including Alibaba for e-commerce, ByteDance for Tiktok social media, Nokia and Baidu.

“Delivering the world’s first 51.2 Tbit/s switch two years after we released Tomahawk 4, the industry’s first 25 Tbit/s switch, is a testament to the outstanding execution and innovation by the Broadcom team,” said Ram Velaga, senior vice president and general manager, Core Switching Group, Broadcom. 

The predictable doubling of bandwidth every 18 to 24 months is a key benefit for these customers, which clearly indicates a 3nm project with foundry partner TSMC.

Six on-chip ARM processors manage the high-bandwidth, fully-programmable streaming telemetry, and provide sophisticated embedded applications such as on-chip statistics summarization.

There is strong motivation to unify the networking infrastructure for general-purpose compute and storage with that of AI/ML compute says the company. AI/ML training clusters are driving the need for fabrics with high-bandwidth connectivity, high radix, and lower job completion time, while operating at high network utilization.

Tomahawk 5 provides AI/ML workload virtualization with features such as single-pass VxLAN routing and bridging. Critical to minimizing job completion time (JCT) for AI/ML workloads, Tomahawk 5 offers features such as Broadcom Cognitive Routing, advanced shared packet buffering, programmable inband telemetry, and hardware-based link failover.

The Cognitive Routing improves network link utilization by automatically and dynamically selecting the most lightly loaded links in the system. This is especially important for AI/ML workloads which often have a combination of short-lived flows and long-lived, high bandwidth flows with low entropy.

Tomahawk 5 includes real-time dynamic load balancing that tracks the utilization of all links, both at the switch and downstream in the network, to determine the optimal path for each flow. It also monitors the health of links in hardware and automatically steers traffic away from failed links. These features provide dramatically improved network utilization and reduced congestion.

Also important is minimizing network congestion by controlling the rate of traffic injected into the network by each source. Since network operators employ a variety of different congestion control algorithms at endpoints such as merchant or custom NICs, Tomahawk 5 provides extensive programmable inband telemetry on both live traffic and network probes.

Real-time metadata can be inserted into traffic at line rate as it traverses the network to collect telemetry on queue size, packet latency, switch utilization, and a variety of other customer-selectable metrics. This metadata can be used for precise end-to-end network congestion control.

To enable the lowest power and lowest cost for physical connectivity, Tomahawk 5 enables a direct 100G PAM4 interface to direct attach copper (DAC), front panel pluggable optics, and co-packaged optics. The flexible, long reach Tomahawk 5 SerDes provides DAC connectivity to all devices within a rack, and even between racks, without the need for retimers or other active components. It can also interface directly to a broad ecosystem of standard front-panel pluggable optical modules.

Tomahawk 5 will also be made available with co-packaged optics using Broadcom’s Silicon Photonics Chiplets in Package (SCIP) platform, providing more than 50 percent decrease in the power needed for optical connectivity. Since the same switch silicon provides all these options, customers can choose the optimal I/O for each part of their intra-cluster, inter-cluster, and inter-DC networks with no software porting required.

“Since the introduction of Tomahawk 1 in 2014, Broadcom has consistently executed on doubling the bandwidth approximately every two years. With today’s introduction of the fifth generation Tomahawk family, we are proud to say that a single Tomahawk 5 replaces forty-eight Tomahawk 1 switches in the network, resulting in over 95 percent reduction in power requirements. We applaud our customers, partners, and engineers for making this possible,” said Velaga.

Along with the Trident and Jericho switch families, the Tomahawk series is part of Broadcom’s three-pronged strategy of providing optimized switch architectures for different network applications. All these devices share a common programming interface, so customers can easily leverage their software development efforts across different platforms.

“Alibaba has deployed multiple generations of Tomahawk-based platforms throughout our network infrastructure. Once again, Broadcom has delivered another robust product in Tomahawk 5, with unmatched bandwidth, radix, power efficiency, and product quality. We are proud to collaborate with Broadcom on Tomahawk 5, leveraging some cutting-edge features such as programmable inband telemetry to build our predictable super performance cluster, and to service the exponential growth of our AI/ML workload and high performance storage requirements,” said Dennis Cai, Head of Network Infrastructure, Alibaba Cloud Intelligence.

“The growth of our short video platforms over the past few years demonstrates the importance of having a scalable, high bandwidth, high performance and efficient network in data centers. Tomahawk 5 will continue to fulfill these metrics, and ByteDance is thrilled to see the readiness of Tomahawk 5 in the merchant silicon ecosystem, which will enable ByteDance to develop its new generation switches and network infrastructure,” said Lark Wang, General Manager of System Department at ByteDanc

“Broadcom’s Tomahawk 5 64x800G switch becomes a valuable tool to help migrate our data center customers into the 800GE era. 800G optics are generally available this year and their use makes sense right away from both a commercial and power perspective compared to 400GE,” said Ken Kutzler, Vice President Hardware Engineering at Nokia.

www.broadcom.com

Related articles

Other articles on eeNews Europe

 

 


Share:

Linked Articles
eeNews Europe
10s