I have spent most of my career balancing two key forces in network decisions. One is the freedom to design a network your way. The other is the discipline required to operate that network at scale, across hardware and software boundaries, over time. In hyperscale environments, teams have already shown that SONiC can deliver on both promises. They have built large, efficient networks around an open NOS and helped move SONiC from an experiment to a proven platform.

Bringing SONiC into the enterprise

The question now is less whether SONiC can scale and more whether it can fit the enterprise. Enterprise networks are not one fleet with one operating model. They bring multiple hardware generations, existing routing and management systems, stricter change controls, and operations teams that may not own the entire software stack. The challenge is bringing SONiC into that environment without creating a new island of operations.

Customers often ask me a similar question: how do we keep the multivendor support, open source, and customizability that made SONiC attractive, while fitting into the support, lifecycle, security, and operational models the enterprise already depends on?

My answer is that it shouldn’t be a trade-off. Flexibility and operational confidence aren’t opposites, and customers shouldn’t have to choose between them.

This is what we’re working on at Cisco. It starts with real, practical choice. Customers can choose the network operating system, hardware, automation, and management tools that fit their needs, and get support from lab testing to production. Right now, there are two main options:

• Build and customize a SONiC distribution from the community, using Cisco platforms as the foundation, with your engineering team managing the stack.
• Run a Cisco-hardened SONiC distribution, with validated releases, Cisco engineering accountability, lifecycle management, and TAC support.

The choice is up to the customer. Our job is to make the two options real, not just something you see in a presentation.

We’re well positioned to do this work. Cisco is a leading member of the SONiC community, with seats on both the Governing Board and the Technical Steering Committee (TSC), and we keep investing in upstream engineering. We have hundreds of thousands of switches with 10 million ports deployed worldwide, and we’re among the top contributors in the community. I share these numbers not to boast, but because we understand what it takes to build and run SONiC in production at large scale.

Openness also opens a door at the silicon layer, which is the part I personally find most fun. With Cisco Silicon One, P4 programmability and SAI provide an open interface between the NOS and the forwarding plane. That interface can support the functions that a data center fabric needs, from IP/BGP, EVPN, and SRv6 to traffic steering, congestion visibility, observability, and telemetry. It gives customers room to evolve forwarding behaviors with the combination of SONiC and Cisco Silicon One without waiting on the next silicon cycle, while Cisco 8000 and N9000 add range across the stack.

Upstream alignment matters, and it’s what I’d focus on if I were buying. Nobody wants to be stuck on a private fork. It can feel faster at first, but it turns into a burden with every release. At the same time, customers can’t always wait for the community roadmap. With Cisco supporting SONiC, customers can come to us when they need a key feature. We can work to deliver it and contribute it back to the community quickly, reducing the risk of permanent divergence. We then validate the release, the platform, and the upgrade path together. The result is speed when customers need it, without trading away the freedom to use their own automation and controllers, and with a smaller gap between what works in the lab and what holds up in production.

Operating and securing SONiC networks at scale

As AI fabrics and data center fabrics scale to tens of thousands of nodes, network visibility and mean time to failure are increasingly important. Traffic patterns change quickly as jobs start, stop, and move. Just seeing a port is up isn’t enough. Operators need telemetry that explains congestion, microbursts, path behavior, and how it all affects the application. SONiC’s open-software model is a strong base for sending this data into the observability and automation systems teams already use.

Interfaces are just as important at the operations layer. OpenConfig models, REST APIs, OpenTelemetry, and network-as-code patterns let teams connect SONiC to controllers, CI/CD workflows, and observability systems. The goal isn’t to make everyone use the same stack. It’s to let teams keep the tools and automation they trust, while giving them a consistent way to manage a multivendor fabric.

None of this works without security. This is the biggest change. As SONiC moves beyond tightly controlled hyperscale environments to shared facilities, remote sites, and edge deployments, security and recovery become especially important. No one may be nearby to fix a failed device. The network needs to establish trust, prove software and device integrity, and support recovery throughout the device’s lifecycle.

With scale comes a broader threat surface. Our customers need to reduce exposure to newly disclosed vulnerabilities while keeping services online and patching on a planned schedule. At OCP Global Summit 2026, Cisco is announcing Live Protect for Cisco-validated SONiC on Nexus One, extending support beyond Cisco NX-OS, where Live Protect is available today. On Cisco N9000 Series Switches, it provides vulnerability shielding on a live system, so operators can protect the network and install patches during a planned maintenance window without downtime or a switch reboot. Live Protect continuously revalidates protection. This support is expected in Q2 2027.

I’ll be covering this progression from hardware trust anchors to software integrity, runtime protection, resilient recovery, and day-two operations. It’s part of Cisco’s broader presence at OCP 2026, where we’ll showcase open networking, Silicon One, AI scale-out, scale-across, optics, telemetry, out-of-band management, and Live Protect on Cisco-validated SONiC so you can see how everything connects from silicon to workload.

If you’ll be there, I hope you’ll stop by. I’ll be speaking on Wednesday, October 14, 2026, at 11:45 a.m. in the Expo Hall Theater at the San Jose Convention Center: “Open, Trusted, and Secure SONiC at Scale.” I will be joined on stage by Dina Papagiannaki, Corporate Vice President of Azure Datacenter & Metro Networking at Microsoft.

Thank you to our SONiC engineering and Silicon One teams for all their hard work, and to the wider SONiC community for keeping the project truly open. I’m looking forward to the discussions. See you in San Jose.

Learn how Cisco is securing and scaling SONiC for modern AI infrastructure at OCP Global Summit.

Read more

• Cisco open networking

• Cisco Silicon One resources

• Cisco Nexus One