Why we replace cabling before it fails, not after

Structured cabling is the one piece of data-centre infrastructure nobody thinks about until it's the reason a rack won't hit its rated speed. Servers get refreshed every 3–5 years. Switches get refreshed on the same cycle. The cabling underneath both often outlives three or four hardware generations — and that's exactly the problem.

What actually degrades

Copper and fibre don't fail cleanly. They degrade gradually, in ways that are easy to miss until they're not:

• Insertion loss creeps upevery time a fibre run is re-patched, re-terminated, or physically disturbed. TIA-568.3-D allows up to 0.75dB per mated connector pair — a connector that tested clean at install can be well past that budget five re-patches later, and a failing run usually fails at the connector, not the cable itself.
• Copper bends past specfrom cable management changes, rack reshuffles, and years of foot traffic in containment routes. The standard minimum bend radius for four-pair copper is four times the cable's outer diameter, during installation and after — not just while it's being pulled. Once that's violated it stays violated until someone re-runs the cable.
• The category ceiling doesn't move.Cat6A does 10G to 100 metres; Cat8 pushes to 40G but only over roughly 30 metres — a different design target, not a drop-in upgrade. Cat5e was never in this conversation. Our cabling partner ONNEC surveyed 250 data-centre decision-makers in 2025 and found 92% had seen AI-driven capacity demand rise over the past year, by an average of 42% — headroom most installs from even three years ago weren't speced to absorb.

The pattern is always the same: nobody notices until a migration exposes it.

The real cost isn't the cable

A structured cabling refresh is a rounding error next to the compute it protects. The expensive failure mode isn't a bad patch cable — it's discovering an undersized backbone mid-migration, during a maintenance window, with a client watching the clock. Reactive replacement happens on someone else's timeline. Proactive replacement happens on yours.

What we actually check

Every install and every recertification we run gets full test and certification documentation to TIA-568.3-D — Tier 1 loss, length, and polarity testing with an OLTS on every fibre link, OTDR (Tier 2) added where the job calls for connector-by-connector fault location, and Fluke DSX/Versiv results for copper. Tier 1 is the mandatory certification test; OTDR traces on their own aren't accepted as certification evidence, so a supplier handing over an OTDR bundle and calling it certified has skipped the test that actually proves the link's loss budget. That distinction is what tells you whether a run is fine, degrading, or already outside spec — instead of guessing from an error rate that showed up three systems downstream.

If you don't know the current state of your cabling plant, that's the actual starting point — not a full rip-and-replace quote. Get an audit before you need one.