Cisco lays groundwork for 800G networks as AI, 5G and video traffic demands grow


Cisco has amped-up its support for 800G capacity networks with an eye toward helping large enterprises, cloud and service providers handle the expected demand from AI, video, and 5G services.

At the core of its recently developments is a new 28.8Tbps / 36 x 800G line card and improved control software for its top-of-the-line Cisco 8000 Series routers.

The 28.8T line card is built on Cisco’s Silicon One P100 ASIC, and brings 800G capability to the modular Cisco 8000 Series Router, which can scale to 230Tbps in a 16 RU form factor with the eight-slot Cisco 8808, and up to 518Tbps in the 18-slot chassis, according to Cisco.

Improving space and energy use while bulking up densities are also part of the new line-card’s story, wrote Satish Surapaneni, director, product management, with Cisco’s Networking, Provider Connectivity group in a blog about the 800G improvements.

The cards can support 36xQSFP-DD800 ports, which can enable the use of 2x400G and 8x100G optics, and deliver densities with 72x400G ports or 288x100G ports per slot. “The reason we can double the density is because the P100 uses state-of-the-art 100G SerDes technology that can achieve higher bandwidth speeds in the same footprint,” Surapaneni stated.

So instead of six 400G modular systems, one 800G eight-slot modular system can achieve 230Tbps with up to 83% space savings, up to 68% energy savings,or about 215,838 kg CO2e/year (green house gases) savings. “To put it in perspective, these carbon savings are the equivalent of recycling 115 tons of waste a year instead of going into landfills,” Surapaneni stated. 

In addition to sustainability and operational cost benefits, customers can protect their pluggable optics investments since Cisco QSFP-DD 800G is backward compatibility to lower-speed QSFP-DD and QSFP modules, Surapaneni stated. 

On the software side, Cisco bolstered its Crosswork Network Automation software that includes tools to monitor and verify network operations, collect and correlate data to identify network trends and patterns, and automate operations. This enables faster provisioning, and provides advanced visualization dashboards, services monitoring, and better network optimization, Surapaneni stated.

Cisco is also adding a new feature to the 8000’s IOS XR operating system. IOS XR Segment Routing innovations with Path Tracing provides hop-by-hop visibility into packets are flow to help detect and troubleshoot issues and promote cost reduction, Surapaneni stated.

Introduced in 2019, the Silicon One family of unified switches and routers is important to Cisco’s goal of network underpinnings of large webscale and service-provider networks. Silicon One optical-routing silicon includes support for large non-blocking distributed routers, deep buffering with rich QoS, and programmable forwarding. The Silicon One family lets Cisco compete with others such as Intel, Broadcom, Juniper, and Arista in building high-capacity networks.

“Given that traffic volumes are increasing, higher capacity is needed at the network intersection points, such as in the core. These core networks are in the IP backbone and metro regions, where we’re seeing more traffic concentrating, as applications and services move closer to the user, user access speeds increase with fiber and 5G, and functionality such as peering, subscriber management and CDN get distributed locally,” Surapaneni stated.

“To avoid traffic jams with network congestion, a scalable metro core is needed to transport all traffic types, particularly high-bandwidth latency-sensitive traffic,” Surapaneni states. “However, metro locations tend to be smaller with tighter space constraints, which is why space efficiency is critical. Scaling to 800G can help providers address space and traffic demands efficiently with metro applications.”

Surapaneni pointed to recent Dell’Oro Group research that stated IP backbones that interconnect metro networks are important to scale and help reduce bottlenecks. According to Dell’Oro, upgrades with IP backbone networks represent the highest demand for 400G since the internet backbone includes both cloud and communications service provider networks that carry mobile, broadband, and cloud traffic. 

“We expect 400Gbps and higher speeds to comprise nearly 70% of data-center switch sales by 2027,” according to a recent reeport, by Sameh Boujelbene, vice president at Dell’Oro Group. “The proliferation of Artificial Intelligence workloads will further accelerate this adoption. Nevertheless, as network speeds increase, pluggable optics may reach physical limitations and no longer satisfy thermal and density requirements.”

But demand for increased capacity is expected to drive 800Gbps to eclipse 400Gbps by 2025. The availability of 800Gbps optics and 25.6T chips propelled adoption of 800Gbps switch ports. This was spearheaded by Google, but other cloud providers are expected to follow suit, Boujelbene says.

Surapaneni stated that as data-center traffic increases, there will be more capacity demands in spine/leaf environments, such as super-spine, in addition to Data Center Interconnect (DCI) and data-center WAN/core networks.

“AI/ML workloads are different from traditional data center traffic because the processors are very high bandwidth devices that can overwhelm networks and impact job completion rates without sufficient spine capacity,” Surapaneni stated.

Copyright © 2023 IDG Communications, Inc.



Source link