What happens, minute by minute, when the fibre is cut
A digger, a trench, and four hours with no link. Here is what stops, what carries on, and exactly when — first on a hundred-per-cent cloud phone system, then in the same company with an Edge on site.
We usually describe an outage as a state: “the line is down”. Seen from operations it is not a state, it is a sequence — and the sequence is the same everywhere, give or take a few seconds. Knowing it changes what goes into a continuity plan, because most plans deal with minute 40 and forget minute 3.
The sequence, on a fully cloud platform
The company has thirty extensions, a voice menu, two call queues and a platform hosted four hundred kilometres away. The digger cuts at 09:14.
- 09:14:00 — the link drops. Calls in progress do not end immediately: the audio goes one-way for a few seconds, and each person hears the other fall silent.
- 09:14:08 — calls in progress are cleared by media-timeout detection. Thirty or so conversations end at the same instant.
- 09:14:30 — the handsets start losing their SIP registration. They display “not registered”, or nothing at all depending on the model, and can no longer dial.
- 09:15 — extension 201 tries to call extension 214, ten metres away. The call fails: it had to go through the data centre.
- 09:16 — inbound calls fall through to whatever overflow is configured at the carrier. At best, a divert to a mobile; in the common case, an engaged tone.
- 09:20 — somebody rings the carrier from their personal phone. A ticket is opened.
- 13:05 — the fibre is spliced. The handsets re-register over three or four minutes, in no particular order.
- 13:30 — the company discovers it has no record of the calls attempted over four hours. They exist nowhere.
The expensive moment is not minute 14. It is minute 15: the point at which two people in the same building can no longer speak to each other, because their conversation had eight hundred kilometres to travel.
The same sequence, with an Edge on site
Same company, same digger, same hour. A Konvoice Edge is installed in the comms room and holds the site’s dialling plan.
- 09:14:00 — the link drops. Internal calls in progress do not move: their media never went over the internet. External calls in progress are lost — that media went through the carrier.
- 09:14:05 — the Edge loses its mTLS link to the cloud, logs the event, and switches to standalone mode. The handsets do not move: they were already registered to it, not to the cloud.
- 09:15 — extension 201 calls extension 214. It rings in under 300 ms.
- 09:16 — an inbound call arrives over the site’s backup line. The voice menu answers, the queue runs, the reception handset rings.
- 09:30 — a caller leaves a voicemail. It is recorded on the Edge, encrypted, and it waits.
- 13:05 — the link returns. The Edge re-establishes it and sends up the logs, the queue counters and the voicemails in the order they happened.
- 13:06 — the dashboard shows the four hours with their real traffic. Nobody has anything to re-enter.
What that difference costs, and what it returns
An Edge is a box, a network socket and a line on an invoice. The fair question is when it pays for itself.
The answer turns on one variable, and it is not the number of extensions: it is whether the site receives the public or coordinates work by telephone. A floor of developers who talk over instant messaging loses very little in four hours. A clinic reception, a loading bay, a high-street agency and an on-call rota lose their day.
Our rule of thumb: past about twenty extensions, or as soon as a site has a physical reception desk, the Edge pays for itself at the first outage. Below that, cloud mode is enough and we say so.
Three questions to ask your current supplier
- “What exactly happens, second by second, when my link goes down?” — if the answer fits in one sentence, it is wrong.
- “Can extension 201 call extension 214 during the outage?” — this is the question that separates an architecture from a brochure.
- “Where are the calls attempted during the outage?” — if they exist nowhere, you will never know what the incident cost.
On that last point, the honest answer for a pure cloud platform is “at the carrier, and you have no access to them”. That is not a failing of the supplier: it is a consequence of where they put the intelligence.
