Data Centre Leaf-Spine Fabric

Data center operations are increasingly defined by the tools and workflows of the compute environment. Ansible orchestrates server deployments, Puppet manages configuration state, systemd governs service lifecycles, and shell scripts automate everything in between. Yet the network fabric that connects these compute resources remains a separate operational domain -- managed through proprietary command-line interfaces, governed by vendor-specific automation adapters, and staffed by specialists trained on platforms that share nothing with the Linux ecosystem surrounding them.

This operational divide creates real costs. Network changes require a different team, a different toolchain, and a different approval workflow than compute changes. Automation pipelines that provision a hundred servers in minutes stall when they reach the network layer, because the switches do not speak the same language as the rest of the infrastructure. Troubleshooting spans two worlds: Linux tools diagnose compute issues, proprietary CLI tools diagnose network issues, and correlating the two requires manual effort and domain-specific expertise on both sides.

Scaling compounds the problem. A single rack with a top-of-rack switch is manageable regardless of toolchain. But as data centers grow from tens of racks to hundreds -- from campus-level deployments to multi-site operations -- the operational overhead of maintaining a separate network management domain grows proportionally. And when the time comes to refresh hardware, proprietary vendors bundle their software with their switches: upgrading the NOS means replacing the hardware, and replacing the hardware means re-qualifying the entire fabric. The vendor lock-in that seemed manageable at rack scale becomes a strategic liability at data center scale.