For thirty cameras you do not need a campus network.

Zyxel puts first solving with little money what a small or medium installation really needs: ports with power for the cameras, separated networks, a fibre link between two cabinets, reasonable Wi-Fi and a screen from which to see all of that without setting up a server. It does not try to compete in the centre of a campus network and there is no reason to ask that of it.

What it sacrifices is depth: less packet memory in the access units, advanced functions in their basic version, network recovery measured in seconds and not in milliseconds, and support that is not that of a top-range house. And the life cycle of a particular model is shorter, which matters when the same switch has to be bought four years later.

It is recognised where most of the real installations are: an industrial unit with offices, a clinic, a school, a small hotel, a large residents' association, a branch office. Twenty, forty, sixty cameras, a recorder, one or two cabinets and nobody from IT inside.

Two paths. The equipment can be configured on its own, from its own web page, the way it has always been done: enough for one cabinet and perfectly valid. Or it can be adopted into the manufacturer's cloud platform and every site administered from one screen, with the devices seen on a map, with the consumption of each port and with alerts when something goes down.

The second way has a level of use at no cost with the basics, and paid levels that add functions and history. That is an important practical difference from the top-range cloud brands: here you can have central management on a small installation without putting an annual fee into a budget that cannot take it.

In the access layer and in the single cabinet of small and medium installations: powering the cameras, separating them onto their own network, closing their way out to the internet, taking the traffic to the recorder and having a fibre link if there are two buildings. That is ninety per cent of what is needed in an installation of this size, and it does it.

And in the place where the budget is real. Part of our work is making the money go where it shows: in an installation of thirty cameras, spending on the right lens, on the right drive and on cabling done properly pays off more than spending it on a top-range switch whose functions nobody is going to switch on. Saying that is part of the work, even if the quotation goes down.

At the point where the traffic of many cameras comes together. There you need packet memory to absorb the hits —every complete frame from every camera is a burst at full speed— and generous uplinks. A cheap access unit at that point drops packets, and what you see is broken image with no error on any screen.

Where the installation cannot go blind even for a few seconds. Path recovery is done with the standard mechanisms and takes of the order of seconds, which with cameras means reconnections and the odd gap in the recording. If that costs money, the ring has to be of another class.

In harsh outdoor conditions, in an industrial plant or in an outdoor cabinet with no climate control. An office switch put inside a car park cabinet restarts on the first hot day, and that goes for every brand: there it has to be industrial equipment.

And in organisations with a demanding security specification, with network access control, change logging and auditing. It is not that the functions are missing: it is that they do not reach the level of detail a specification like that asks for, and forcing it is signing something that cannot be demonstrated.

The equipment is bought and it works. The cloud management has a basic level at no cost and paid levels per device that add functions, alerts and more history. And on the firewall side, as everywhere in the industry, the security services —content filtering, signatures— go by subscription.

When a paid level expires, the switch does not stop: it carries on switching, it carries on giving power and it can still be administered, although with fewer functions and less history. That is the underlying difference from the brands where the licence is what keeps the equipment switched on, and in a small installation that difference is peace of mind: if one year nobody renews anything, the surveillance carries on recording.

If there is somebody from IT —in-house or external—, the network is theirs and we ask for what we need in writing: this network for the cameras, these watts, this throughput, this can talk to this and nothing else. It is the usual conversation and it is settled in half an hour when you turn up with the sum done.

If there is nobody, which is frequent at this size, then the installation has to be left ready for a maintenance person with no network training to run it. That is three specific things: the cloud management screen in the customer's name with their own account, a one-page procedure with the four actions they can take —see which camera is down, restart a port, check a cable, call— and the map of which camera is on which port with which address, printed and stuck inside the cabinet.

What we are not going to do is keep the switch password so that they depend on us. The administrator credentials go in the customer's name. If they want to call us, let it be because they find it useful and not because they cannot get into their own network.

And the tone, which runs through all ten: if there is already a network built and somebody runs it, we adapt to it. We only propose changing something when it genuinely does not reach —the power, the throughput, the separation— and we explain it with numbers before signing, not in the third month.