Office networks tend to grow by accretion. A switch here, a consumer router there, a run of cable to the new desks. It works until roughly the thirtieth person, and then it produces intermittent faults that are extremely difficult to diagnose.
Decisions worth making deliberately
- Structured cabling over daisy-chained switches. Cable is cheap while the office is empty and disruptive once it is full
- Business-grade access points rather than consumer routers. The difference shows when forty devices are connected, not four
- Separate networks for staff, guests, and any card payment or CCTV equipment. Segmentation costs nothing at setup and contains problems later
- Static addresses for anything other devices need to find — printers, servers, network storage
- An uninterruptible power supply for network equipment, not only for servers. A router that reboots on every outage looks exactly like an internet fault
Write it down while you still remember
The single most valuable artefact is a current diagram: what connects to what, which port, which address range, which credentials. It takes an hour during installation and saves days two years on, when the person who set it up has moved on.
Photograph the rack and the patch panel labels. Keep the diagram somewhere other than that same server.
The pattern behind most complaints
When people say the internet is slow, the cause is usually inside the building: an overloaded access point, a duplex mismatch, one machine saturating the uplink with an unattended sync. Without monitoring you cannot tell, so you escalate to the provider and lose a week.
Basic monitoring on the gateway — throughput and uptime, nothing exotic — converts that week into a ten-minute answer.
- Infrastructure
- Hardware