November 22, 2026

Bridging the Gap Between Cloud-Native and Core Networking: Mauricio Rojas on NetOpsKube and the GitOps Revolution

The landscape of network operations is undergoing a seismic shift, moving away from the traditional, manual command-line interface (CLI) approaches of the past toward a future defined by automation, programmability, and cloud-native principles. At the heart of this transformation is the integration of Kubernetes and GitOps into the networking stack—a transition that promises to bring the agility of software development to the traditionally rigid world of hardware infrastructure. In a recent and deeply technical exchange, Eric Chou sat down with Mauricio Rojas, a Principal DevOps Consulting Engineer at Nokia, to explore how these technologies are converging to redefine modern network management.

The conversation centered on the practical application of these tools, specifically focusing on Rojas’s pioneering work with the NetOpsKube project. As networks become increasingly complex, the industry is looking toward open-source tooling and standardized orchestration to manage everything from core aggregation to the edge. The dialogue provided a comprehensive look at how the principles that revolutionized the data center are now being applied to the very backbone of the internet, including Broadband Network Gateways (BNGs) and core routing infrastructure.

The Evolution of NetOps: From Manual Configuration to Kubernetes Orchestration

For decades, network engineering was a discipline of individual expertise and manual intervention. Engineers would log into devices one by one, applying configurations that were often documented only in static text files or spreadsheets. However, as the scale of global networks has exploded, these traditional methods have become unsustainable. Mauricio Rojas, through his role at Nokia and his contributions to the networking community, has been at the forefront of advocating for a "NetOps" approach that mirrors the DevOps movement in software engineering.

The core of this shift is Kubernetes. While originally designed for container orchestration in the application layer, Kubernetes has proven to be a remarkably flexible platform for managing almost any declarative state. In the context of networking, Kubernetes serves as the control plane that ensures the actual state of the network matches the desired state defined by the engineers. This move toward a "controller-based" architecture is a departure from the "script-based" automation of the early 2010s, providing a more robust and self-healing framework for infrastructure management.

Rojas’s project, NetOpsKube, represents a tangible manifestation of this philosophy. By leveraging a stack of Kubernetes, GitOps workflows, and specialized open-source tools, NetOpsKube provides a blueprint for core and aggregation operators to manage their networks at any scale. The project aims to democratize high-end automation, making it accessible not just to hyperscalers with massive engineering teams, but also to smaller regional providers and enterprise operators who need reliable, repeatable, and transparent network management.

NetOpsKube and the Role of GitOps in Modern Infrastructure

A significant portion of the discussion focused on the mechanics of GitOps and its necessity in modern networking. GitOps is a methodology where the entire configuration of the infrastructure is stored in a version-controlled repository, typically Git. This repository serves as the single source of truth. When a change is needed, an engineer submits a pull request; once reviewed and merged, an automated system—often a Kubernetes operator—synchronizes those changes with the physical or virtual network devices.

This approach brings several immediate benefits to network reliability. First, it provides a complete audit trail of every change made to the network, including who made the change and why. Second, it allows for rapid rollbacks in the event of an error. If a configuration update causes an outage, reverting to the previous known-good state is as simple as reverting a commit in Git.

In the NetOpsKube ecosystem, this is made possible through tools like the gNMIc Kubernetes Operator. The gRPC Network Management Interface (gNMI) is a protocol developed by Google and now widely supported by major vendors like Nokia, Arista, and Cisco. It allows for the streaming of telemetry and the pushing of configurations in a structured, model-driven format. By wrapping this capability in a Kubernetes operator, Rojas has enabled a system where Kubernetes can "watch" the network configuration and automatically correct any "drift"—the phenomenon where the actual state of a device deviates from the intended configuration due to manual intervention or hardware resets.

Transforming the Edge: BNGs and Subscriber Management

One of the more specialized areas Rojas and Chou explored was the application of these cloud-native principles to Broadband Network Gateways (BNGs). The BNG is a critical component in service provider networks, acting as the point where individual subscribers connect to the provider’s backbone. It handles essential functions such as authentication, authorization, and accounting (AAA), as well as traffic shaping and policy enforcement.

Historically, BNGs have been some of the most complex and difficult-to-automate components of a network because they deal with dynamic subscriber sessions rather than just static routing tables. Rojas discussed how the move toward Disaggregated BNGs (DBNG) and the use of the Control and User Plane Separation (CUPS) architecture allows for more granular management. By applying a NetOps stack to these environments, operators can manage subscriber policies and gateway configurations with the same level of automation used for web servers.

This level of automation is particularly relevant as operators look to scale their broadband offerings. The ability to deploy new BNG instances or update subscriber policies across a wide geographic area without manual intervention is a significant competitive advantage. It reduces the time-to-market for new services and minimizes the risk of human error in high-stakes environments.

The Intersection of AI, LLMs, and Network Operations

The conversation naturally pivoted to the most discussed technology of the current era: Artificial Intelligence. Specifically, Rojas and Chou examined the role of Large Language Models (LLMs) and the emerging Model Context Protocol (MCP) in the future of network operations.

While AI has been used in networking for years—primarily for anomaly detection and predictive maintenance—the rise of LLMs introduces a new interface for network management. Rojas noted that LLMs can assist engineers in generating complex configurations, interpreting dense technical documentation, and even suggesting troubleshooting steps based on telemetry data. However, the challenge remains in ensuring that these models have the correct context and that their outputs are safe to implement on live production networks.

This is where the Model Context Protocol (MCP) becomes vital. MCP is designed to provide a standardized way for AI models to interact with external data sources and tools. In a networking context, this could allow an LLM to safely query the current state of a router or check the Git repository for recent changes before suggesting a fix. By grounding the AI in the actual state of the network, operators can mitigate the risk of "hallucinations" and ensure that the AI’s suggestions are technically sound and compliant with organizational policies.

Addressing the Security Challenge: Managing CVEs in the AI Era

As automation and AI increase the speed of network operations, they also introduce new security challenges. The sheer volume of software and firmware updates required to keep a modern network secure can be overwhelming. This issue was highlighted by the discussion surrounding Gluware, a leader in intelligent network automation.

One of the primary concerns for modern operators is the "deluge of CVEs" (Common Vulnerabilities and Exposures) that are increasingly being identified, often with the help of AI-driven security research. Managing these vulnerabilities across a fleet of thousands of devices is a monumental task. Gluware’s Titan Exposure Management was noted as a solution that addresses this problem by assessing every known CVE against the specific hardware, software version, and—crucially—the operational state and configuration of network devices.

In many cases, a vulnerability may exist in the software of a device but may not be exploitable due to the specific way the device is configured. Traditional vulnerability scanners often miss this nuance, leading to a high number of false positives that waste engineering time. By integrating configuration and operational state into the assessment, tools like Titan allow engineers to prioritize the most critical threats and implement compensating controls or automated patches where necessary. This capability is an essential component of a modern NetOps strategy, ensuring that security keeps pace with the speed of automated deployments.

Community Collaboration and the Road Ahead

The insights shared by Mauricio Rojas are not just theoretical; they are being actively discussed and refined within the global network engineering community. A significant milestone for the NetOpsKube project was its presentation at NANOG 97 (North American Network Operators’ Group), where Rojas delivered a tutorial on building a NetOps stack for operators of any size.

The NANOG community has long been the proving ground for the technologies that run the internet. The interest in Kubernetes-based network management at such a forum signals a broader industry acceptance that the "old ways" of networking are giving way to a more integrated, software-defined approach. The availability of the NetOpsKube repository on GitHub further encourages this collaborative spirit, allowing engineers to experiment with the stack and contribute back to the project.

As the interview concluded, the overarching theme was one of cautious optimism. While the transition to a Kubernetes and GitOps-based network stack requires a significant shift in mindset and a commitment to learning new skills, the rewards are clear. By adopting these cloud-native principles, network operators can achieve a level of stability, scalability, and security that was previously impossible.

The journey toward a fully automated, self-healing network is still in its early stages, but with projects like NetOpsKube and the continued evolution of protocols like gNMI and MCP, the roadmap is becoming clearer. For engineers like Rojas and Chou, the focus remains on building the tools and the community necessary to support the next generation of global connectivity. As the lines between the data center, the cloud, and the network continue to blur, the principles of NetOps will undoubtedly become the standard by which all modern infrastructure is measured.