Cisco powers up Nexus switch, offers 800GB optic modules

Cisco is using its high-powered Silicon One chip technology to turn up the power and efficiency of its Nexus family of data center, hyperscaler and cloud switches.

The company rolled out a new high-end Nexus switch for the data center and one aimed at disaggregated applications. Cisco also added an 800Gb Ethernet module. Each of the new additions is powered by the company’s advanced Silicon One technology.   

Introduced in 2019, Cisco’s Silicon One architecture uses the vendor’s custom chip technology, which features optical-routing silicon, deep buffering with rich QoS, and programmable forwarding.

Silicon One boxes are programmable and can be customized for a range of applications from a single chipset, eliminating the need to deploy multiple, specific silicon for standalone processors, line-card processors, and fabric elements, according to Cisco. This is accomplished with a common and unified P4 programmable-forwarding code and SDK, Cisco says.

Since introducing the portfolio, Cisco has expanded the Silicon One family from large, hyperscaler- and telco-targeted applications to its Catalyst family and now the Nexus line.

Cisco added a new high-end box, the Nexus 9232E/Cisco 8111. This 1RU box features a single 26.5T Silicon One G100 processor chip and can be configured to support 32 ports of 800G, including 64 ports of 400GbE or 256 ports of 100GbE using breakout cables to enable significantly higher densities in a small footprint.

Cisco has 400GB Nexus 9300/9500 boxes shipping today.

The 9232E is aimed at large enterprises, hyperscalers and cloud providers and runs the Nexus NX-OS, while the Cisco 8111 is aimed at telcos or others that want to run disaggregated systems and will run SONiC or other non-Cisco network operating systems.

There are multiple use cases for this box, with AI/ML, video, financial and enterprise edge leading the way, according to Gurudatt Shenoy, vice president of product management with Cisco’s Mass-scale Infrastructure Routing Group.

“There’s a lot of demand from large enterprises as well as webscalars around building AI/ML networks for their own use cases at the back end or other applications that require high-density capabilities,” Shenoy said.

AI/ML capabilities are crucial for digital-twin type predictability of network change results, as well as for modeling highly customized application outcomes in the real and meta world. By 2025, 44% of global data created in the core and edge will be driven by analytics, AI, and deep learning, and by an increasing number of IoT devices feeding data to the enterprise edge, Cisco stated in a blog about the announcement.

Additionally, a microservices architecture is the preferred model for app development today, which has created a diversity of workload deployment models including on-prem, virtual machines, containers, bare metal, serverless, etc. In turn, this explosion of workloads has created more traffic within and among data centers and clouds than ever, Cisco stated.

Further, Cisco noted, these trends are enabled by the speed and scale of bandwidth evolution, including 400G technology. According to Dell’Oro Group, sales of 400G technology are expected to increase 224-fold from 2018 to 2026, while 800G technology will see an 84-fold uptick.

Power reduction and space requirements are also part of the Silicon One G100 story.

Cisco says compared to networks built with 12.8Tbps switches, the G100-based systems can provide 25.6Tbps with up to 77% reduction in power and 83% reduction in space required, and they can save nearly 9,000 kg of CO2 emissions per year.

“Minimizing the power and space these high-density data centers require are some of the primary goals with the G100,” Shenoy said. “Minimizing the form factor and supporting 25.6T capacity in the G100 addresses those issues.”

Furthering the form factor initiative, Cisco rolled out two new high-density, industry-standard QSFP-DD800 optical transceivers that double the port bandwidth and connect single mode fiber links in the data center up to 2km. These new optical transceivers let customers maximize bandwidth on new 800G platforms with high density breakouts to 400G and 100G interfaces while protecting existing investments with backward compatibility to QSFP transceivers, Shenoy said. The new transceivers will be available in Q1 2023.

The anticipated shipping date for the Cisco Nexus 9232E is in first half of 2023.