At some point, most networks stop being designed. They start being accumulated. New switches get added. Old ones stay in place longer than they should. Cabling evolves over time. Eventually, you end up with something that works… …but doesn’t behave predictably.

That’s where this project started. On paper, it was simple:

replace old, out-of-support switches

In reality, it was something else. From the plan:

“redesign the network at layer 1… using a template based design so that we can predict how the switches… will react in failure scenarios”

That was the goal. Not just new hardware. Predictability.

Each SER (Switch Equipment Room) had its own quirks.

  • Different cabling layouts.
  • Different assumptions about which ports connected where.

It all worked — but only because people understood it. The problem is, that kind of understanding doesn’t scale.

So we introduced templates.

A consistent way of wiring switches:

  • defined uplinks to the core
  • defined inter-switch connections
  • clear separation of roles

Once that exists, other things become easier. You can:

  • prefer certain links for specific VLANs
  • understand traffic paths
  • predict behaviour under failure

It’s the difference between:

“it should be fine”

and

“we know how this will behave”

The work itself was physical.

  • Out-of-hours switch replacements.
  • Disconnecting and reconnecting cables.
  • Labelling everything properly.

That last part turned out to matter more than expected.

From the plan:

“+30 minutes per SER… due to cable labelling”

That wasn’t a mistake. It was a correction. The first deployments revealed the reality of the environment. So the plan adjusted.

That pattern repeated throughout:

  • build in the lab first
  • deploy in stages
  • learn from each step
  • update the approach

Nothing about this project was particularly complex.

  • No new protocols.
  • No cutting-edge technology.

But it changed something important. The network became understandable. Not just to the people who built it, but to anyone who needed to work on it later and that’s often the real goal. Not just making something work. But making it predictable.

Tags: