While building our new network backbone, we tested five newly constructed darkfibre routes (and the DWDM systems on top of them) before putting anything into production. The carrier supplied measurement protocols for every route, and in all cases they looked clean.

Only two of the five routes worked flawlessly right away. The other three had issues:

  • Two routes with incomplete cabling in the data centers. The fibers simply ended after about 170 meters inside the Meet-Me-Room, because the last patching hadn’t been done.
  • One route with excessive attenuation on a single fiber. After cleaning the connectors, the fiber was back within spec.

None of this was visible in the carrier’s protocols.

Why the Protocols Look Fine

The carrier only measures its own segment. The cross-connects (XCs) inside the data centers are not part of their measurement at all. The problems we found sat exactly at the boundaries between suppliers: the Meet-Me-Room, the cross-connects and the patching between the carrier and the data center operator. Every party can hand over a correct protocol for its own part while the end-to-end path is still broken.

What We Do Now

We run our own end-to-end measurements on every route before accepting it:

  • OTDR from both ends, to locate events such as connectors, splices and unexpected fiber ends, and to compare the real distance with the expected one.
  • Light source and power meter for the end-to-end insertion loss, which is the number the optics actually care about.
  • Connector inspection and cleaning before every measurement. Dirty connectors are by far the most common cause of unexplained loss.

An OTDR trace that ends at 170 meters instead of several kilometers is unmistakable, and it takes minutes to find, compared to debugging a dead DWDM link later. Discussions around the original LinkedIn post also pointed out two useful additions: running the OTDR with and without filters in DWDM environments, and validating the line with a spectrum analyzer.

Takeaway

Carrier protocols are a good starting point, but they don’t replace your own acceptance test. Measure end to end, keep the traces as a baseline for later troubleshooting, and clean before you blame the fiber.