Avatar

Back in December of 2019, we took a bold step when we announced that Cisco would be evolving from a pure full systems company to one that also embraces a true disaggregated consumption model for the open-source community. Customers can now leverage our world-class technology in a way that best meets their needs. Our entrance into the routing and switching merchant silicon market is a unique value proposition. Cisco stands alone in this space because not only do we consume our own technology in products like the Cisco 8000, but we also offer that same technology to external customers.

We knew we had to do this right the first time. Having a competitive silicon device and roadmap available to the broad market is one thing but delivering it with a complete tool kit with expert support is critical to complete the equation for all phases of customer development.

Figure 1. Cisco Silicon One Consumption Models

 

At the time, many people assumed Cisco wasn’t serious about being a component supplier, but today we’re pleased to publicly announce, Meta (formerly Facebook) has deployed the Cisco Silicon One Q200L device along with the Wedge400C Top of Rack (TOR) switch.

As part of the growing Silicon One product family, the Q200L announced in October 2020 is a 7nm device that provides ultra-low power, high performance, a fully unified packet buffer, and the industry’s only fully converged silicon architecture applicable across web scale switching and service provider routing use cases. With 12.8Tbs of total capacity, the Q200L supports up to 16 ports of QSFP-DD56 and 32 ports of QSFP56, providing a combination of 100Gbps, 200Gbps, and 400Gbps rates.

As both a major consumer and developer of silicon technology ourselves, we knew from the very beginning we couldn’t simply ship off sample devices, email a datasheet, and wish our customers luck. Over the years, we’ve learned a few things about building highly available and reliable systems platforms and the spirit of partnership it takes with the component ecosystem. Now, as a component supplier ourselves, we have brought decades of this systems expertise to partner with Meta’s talented and highly experienced engineers on the Wedge400C from both a hardware and software perspective.

Our systems design team worked with Meta’s exceptional hardware team as well as their Original Design Manufacturer (ODM) to successfully support the design. In parallel, our software development team supported the integration of the Cisco Silicon One architecture, armed with our complete Software Development Kit (SDK) and the P4 open-source programming ideally suited for open-source environments like Facebook Open Switching System (FBOSS). Additionally, our implementation of a Switch Abstraction Interface (SAI) gives our customers the ability to achieve complete operating system portability by largely decoupling them from vendor-specific firmware.

Meta’s hardware and software teams, in conjunction with their ODM, built a system using Cisco Silicon One running their FBOSS operating system which can adapt to their evolving use cases.

Figure 2. Wedge400C and Cisco Silicon One

Our Cisco team was honored and thrilled to work with Meta’s world-class team to bring this program to life. The teams quickly began working together as one virtual team across companies, time zones, and geographies towards a common goal of bringing Wedge400C to production. It was a great example of how companies can partner together to break down the traditional barriers and challenge conventional thinking. With Wedge 400C now in full production, it is helping facilitate data center efficiency with best-in-class power, programmability, and performance.”

Stay tuned for new developments! In the meantime, take a few minutes to learn more about Cisco Silicon One architecture, devices, and benefits.



Authors

Rakesh Chopra

Cisco SVP & Fellow

Common Hardware Group Architecture and Platforming