I recently came across an insightful blog by Bruce Davie in The Register that sheds light on the advancements in virtual network automation. While his analysis of virtual networks is insightful, he raises an important question about the state of physical network automation, “Are we actually any closer to the automation of networking than we were a decade ago?” My answer is a definitive ‘Yes’.

Back in 2014, when network automation for physical networks was still in its infancy, we faced a steep learning curve. Learning resources were limited, and most progress was made through hands-on experimentation. When I began my NetDevOps journey at my previous company, IPsoft, a managed services provider, our initial efforts were focused on basic automation tasks, such as connecting to network devices via a shell script and optionally issuing a string of commands. Despite its simplicity, it had a significant impact when scaled. By streamlining the login and information collection process, we saved our team hundreds of hours each month. This early success laid the groundwork for more sophisticated automation developments that followed.

Over the next few years, our automation capabilities evolved significantly. We developed an in-house automation platform that utilized Perl scripts and a list of devices to handle extensive network changes. This innovation drastically improved our efficiency, reducing project timelines from weeks to days. By automating nearly all change tasks, from interface updates to operating system upgrades, we not only saved time but also enabled our engineers to concentrate on more impactful work. Another notable example was a rudimentary low-code solution that empowered our network engineers to create basic automations and triggering mechanisms. As the team got more comfortable with programming concepts, a customizable widget using javascript logic was incorporated to enable greater variation and refinement in the low-code solution. By harnessing the creativity of the network engineering team and creating a shared environment where automations were exposed and built upon, automations were able to reduce the level of effort for many customer interactions and incident responses by over 90%. These shifts underscored the transformative power of equipping network engineers with DevOps skills.

In 2018, I transitioned from IPSoft to Cisco’s Managed Service Department, where I embarked on the next phase of my network automation journey. There, we confronted a significant challenge: A large portion of our staff was bogged down by repetitive incident management tasks. After analyzing the incidents, we discovered that 10-15 workflows dominated our workload. By automating these workflows, we achieved substantial time savings. Over the next eighteen months, our efforts resulted in 24% of tickets being resolved automatically, 25% were enriched with diagnostic data, and an astonishing 95% improvement in resolution speed was achieved. Our solutions managed to handle over 310,000 tickets, out of a total 1.2 million tickets, without human intervention, again demonstrating the potential of automation in physical networks.

In 2022, I joined Network to Code, where network automation is central to the business model. This role exposed me to some of the most significant advancements in the field. Our open source platform, Nautobot, services as a Network Source of Truth and automation platform. By integrating Nautobot with disparate sources of record, an orchestration system, an automation engine and a telemetry and observability stack with the network infrastructure, new and amazing possibilities emerge. For instance, this architecture allows for the adoption of intent based networking and the maintenance of nuanced configuration standards across the entire network. When automations discover discrepancies between the intended and actual state of the network are identified, they can draft and recommend resolutions or execute them autonomously. This solution significantly reduces configuration drift, the frequency and duration of network anomalies, and minimizes the attack surface of the network. Additionally, this architecture enables the automated deployment of new sites, racks and devices based on standardized deployment models. The solution programmatically designs and configures new deployments, automatically updates IT management systems and pushes fully functional configurations to end devices.

While the field of network automation has seen impressive progress, there are still notable challenges. The integration of NetDevOps practices into network engineering has been slower than anticipated. In their recent 2024 Hype Cycle for I&O Automation, Gartner stated “We estimate that less than 10% of enterprises actively use NetDevOps practices currently. Nevertheless, this trend is changing rapidly due to key industry efforts. Cisco’s introduction of DevNet certification in 2020, which saw over 10,500 engineers certified in its first year, is a testament to the growing emphasis on these practices. Additionally, while traditional on-premises network automation can be demanding and resource-intensive, recent advancements in cloud-based solutions present a more accessible alternative, drastically reducing the setup and maintenance effort.

The past decade has witnessed significant advancements in automating physical networks. What began with basic scripts has evolved into sophisticated architectures that have greatly improved network management efficiency.  Although challenges remain, the advancements in automation and the increasing adoption of NetDevOps practices point to a promising future for network operators. As organizations continue to integrate these technologies, they can look forward to enhanced productivity, reduced costs, and improved network stability, scalability, and security. This progress highlights a broader shift towards smarter, automated solutions that are set to redefine network management for years to come.