Your network is slow, and it's probably not the fibre

When someone says the network is slow, they are usually pointing at the wrong thing. Optical fibre adds about five microseconds of delay per kilometre — across a whole industrial park, that is nothing. So "the fibre is slow" is almost never the fibre.

What people report is a symptom. What you are looking for is a cause, and it is usually somewhere else entirely: a dirty connector, a failing transceiver, a switching loop, a duplex mismatch, a port on the wrong VLAN.

Measure, don't guess#

The habit that saves the most time is refusing to act on suspicion. Before touching anything, I answer two questions:

• What is the scope? One PC, one building, or the whole park? That alone rules out half the possibilities.
• Is it steady or bursty? A continuous ping to the gateway shows whether the loss is constant or only spikes under load.

ping 10.10.0.1 -t

If that ping is mostly 1–2 ms with the occasional 350 ms spike and a timeout, that is not an application failing — that is a link shedding frames.

Work up from the bottom#

From there it is the physical layer first. The switch's own interface counters — CRC errors, receive drops — point straight at Layer 1 the moment they start climbing. Optical power tells the rest:

A healthy receive level sits around −8 to −15 dBm. A reading of −20 dBm is a light budget quietly falling apart — often just a dirty LC connector.

More often than not, the fix is a thirty-second clean with a one-click cleaner, and the "software bug" everyone chased for a week turns out to have been a smudge on a fibre end face.

The takeaway#

Diagnose by layer, measure at every step, and let the readings — not the loudest complaint — tell you where the fault is. By the time you actually touch the fibre, you should already know it is the fibre.

More on the tools and the harder faults — broken fibre, OTDR traces, spanning-tree loops — is in the case study.