So I'm finishing up my master's thesis and I'm kind of stuck on something that's been bothering me for a while.
For my project I built a full automation pipeline for a VXLAN/EVPN data center fabric running Nokia SRLinux and Arista together in ContainerLab. The idea is NetBox as source of truth, Ansible pulls from it and generates configs through Jinja2 templates, validates them against YANG models, then pushes via NETCONF. GitLab handles the CI/CD and version control so every change is tracked and tested before touching the network. Next step is integrating pyATS to spin up a validation lab in ContainerLab automatically and run network tests before anything reaches the main topology. The interesting part is doing all this across two vendors simultaneously.
The problem is every paper and book I read keeps talking about network automation in the context of SDN, like automation only makes sense if you have a controller somewhere. But that's not what I built at all and honestly not what I see people actually using in a lot of environments.
My current thinking is that what I built is kind of a parallel approach to something like Cisco ACI — same end goal of having a programmable automated network, just without the proprietary controller in the middle. And for multivendor environments specifically it actually makes more sense because no single SDN controller really handles Nokia and Arista together properly anyway.
But I'm not sure if that framing makes sense to people who actually work in networking or if I'm missing something obvious.
So a few questions if anyone has time:
- Do you actually see SDN controllers deployed that much in production or is it mostly Ansible/NETCONF type automation?
- Is the multivendor thing a real pain point or am I overblowing it?
- Would you consider this kind of pipeline a realistic alternative to something like ACI for a mid-size DC?
- How would you position this kind of work academically relative to SDN?
Appreciate any thoughts, even just a sentence or two helps honestly