VodafoneZiggo.nl uses cookies, by using this site you accept our cookie policy.

Playing table tennis with data packets

How we keep our network in top shape using data

There are about 500,000 kilometres of cables in the VodafoneZiggo network, so it’s not easy to troubleshoot problems; especially if it’s done manually. VodafoneZiggo likes to use big data to find problems, preferably before the customer is even aware of them. Better still; before they actually happen. HFC Network Manager Huib van Tilburg reveals how data can help with maintenance.

If you’re a VodafoneZiggo customer, you may well have noticed the green cable in your meter box. It passes through a transfer point to an end amplifier box in your street, where it’s bundled with cables from your neighbours' networks. These cables are then connected with other bundled cables in a group amplifier box, and then connected to a fibre node. This node has a fibre connection to our core network; from a local data centre to a regional hub, to the national station, where we connect to the rest of the world... are you still following us? You wouldn't be the first to get confused!

Noise on the line
Technicians also find it difficult to locate a problem among this mass of cables. ‘And the problems don’t end there,’ says Huib. ‘If an excavator accidentally cuts a cable in half, we immediately know the cause of a problem. However, most problems are much more subtle and arise sporadically. For example, a connection in a cabinet might not be quite tight enough, and when a truck drives past, the signal is temporarily lost; or the opposite, the problem is temporarily resolved! The weather also affects our network. Components expand in the summer heat, making it easier to find problems. In the winter they shrink, which makes detection more difficult.’

The new way of looking into the future
‘My Operations team and I use smart software and big data to find faults more easily. We can work miracles through our customers' modems. Mind you, we do absolutely nothing that compromises customer privacy, and keep well away from personal data. So, what do we do? We play a kind of table tennis with all the modems in an area, except we use packets of data instead of a ball. If your modem returns this data slower or less completely than your neighbours’ modems, we know that something is wrong at your end. If we find the same problems with your neighbours, we know that we’re dealing with a problem in the network itself.’

Prevention is better than cure
So we play ‘table tennis’, street by street, neighbourhood by neighbourhood, to detect the source of problems. Our self-learning software detects the tiniest issues. It can reveal a problem, or even where a problem is about to arise. This enables us to intervene before anything actually happens. Some simple numbers show how skilled we are in this; Only 0.1 per cent of all problems on our network are reported by customers. So we find 99.9 percent of problems ourselves, and deal with them before anyone notices.’

Measurements via modem
Previously, cable damage was only detected by reacting to reports. ‘A customer would have to experience a problem and report it. A technician would then visit the customer with a measuring device, with which they sent a signal through the cable. If the signal was reflected, it meant something was wrong. Nowadays, our modems can perform these measurements independently and detect any damage to the cable between the modem and the cabinet in the street. These tests allow us to remotely detect very minor damage, well before any real failures arise. This enables us to predict problems remotely and send a technician before a customer calls.’

Even more precise
Huib's team is already highly successful in detecting problems in certain parts of the network, but he highlights two areas in which the team wants to improve considerably. ‘Firstly, we want to be able to find out with even more precision which part of the route between the modem and the national station has a problem. Secondly, we also want to identify the underlying cause of the problem. That’s the key to making our predictive maintenance really complete.’