Flaky Wi-Fi and a Flat Network in a Three-Storey Office
A 120-staff office occupies three floors of a building. All devices sit on a single 192.168.1.0/24 network served by six unmanaged switches daisy-chained from floor to floor, with one 1 Gbps fibre link to the ISP. Twelve IP cameras and a VoIP handset on every desk share that same network. Wireless access points on every floor are set to channel 6 and transmit at maximum power. Users report that Wi-Fi drops between 10:00 and 11:00 and again at 14:00, that VoIP calls become choppy when large files are copied, and that the camera feed freezes during those peaks. Re-cabling the building is not funded this year, but the network team can reconfigure and add a small number of devices.
1. Should the team keep the flat /24 and add bandwidth, or segment the network? Justify with an address-count analysis and the evidence in the scenario.
Read the worked answer
Segment. Adding bandwidth would not fix the reported symptoms.
Address count. A /24 gives 254 usable hosts. With 120 desks, roughly 120 VoIP handsets (one per desk), 12 cameras, printers, APs, servers and guest devices, the office is at or past the ceiling. Running out of addresses produces exactly the intermittent failures described: clients that cannot get a DHCP lease appear as 'Wi-Fi drops' and recover later.
Broadcast and contention. One broadcast domain carrying cameras (constant multicast and large frames), VoIP (latency-sensitive, small frames) and bulk file copies means a large copy on one floor queues voice packets behind it. That is the stated cause of choppy calls. More fibre capacity upstream does nothing about the shared switch fabric and broadcast domain inside the building.
Proposed segmentation using the existing cabling:
- VLAN 10 - data/desk, 192.168.10.0/25 (126 hosts, with headroom)
- VLAN 20 - voice, 192.168.20.0/26
- VLAN 30 - cameras, 192.168.30.0/27, isolated from desk traffic
- VLAN 40 - guest Wi-Fi, 192.168.40.0/26, internet only
- VLAN 99 - management
This requires at least one managed Layer 3 switch or router-on-a-stick, not a full recable. Inter-VLAN routing with ACLs keeps cameras off the desk VLAN, which is both a performance and a security gain.
2. Diagnose the Wi-Fi peak-hour drops and specify concrete configuration changes, distinguishing what you would change first and why.
Read the worked answer
Diagnosis. All APs on channel 6 at maximum power is co-channel interference by design: every AP hears every other AP and clients associate with a distant, strong AP rather than the nearest one. The 10:00 and 14:00 peaks are simply when the highest number of clients are active, so the contention becomes visible.
First changes (no spend)
- Set a non-overlapping 2.4 GHz plan: 1, 6, 11 assigned so adjacent floors differ. With three floors, alternate floors use the same channel only if physically shielded; vertical bleed-through usually means floor 1 and floor 3 can share.
- Reduce transmit power to medium/low and enable band steering to 5 GHz. Lower power is counter-intuitive but shrinks each cell so clients roam to the correct AP.
- Enable a minimum basic rate so distant clients cannot hold the cell at 1-2 Mbps.
Why this order. Channel plan and power are free and reversible in one maintenance window, and they address the mechanism. Buying more APs first would add more co-channel talkers on the same channel.
Second wave if measurements still show drops - add one AP only where a site survey shows a coverage hole, and confirm the DHCP scope in the voice VLAN is large enough for peak handsets, since a starved scope also presents as 'Wi-Fi drops'.
3. The camera feed freezes during the same peaks. What would you check, and how would you sequence the fix against the VoIP problem?
Read the worked answer
Check in this order
- Uplink saturation between floors. Daisy-chained unmanaged switches mean all camera traffic and all desk traffic share one inter-switch link. If 12 cameras stream at 4 Mbps each, that is 48 Mbps of steady traffic before any file copy, on a link that may be negotiating at 100 Mbps rather than 1 Gbps due to a cable fault. A duplex mismatch or a 100 Mbps link would explain freezes precisely at peak.
- Camera recording target. If the NVR is on a different floor from the cameras, every stream crosses the daisy chain twice.
- Power. If cameras are PoE and the injector or switch is near its power budget, cameras reboot or stall.
Sequencing. Fix VoIP first: it is real-time and cannot retry, whereas video can buffer and is recorded locally. Practically this means prioritising voice in QoS and putting cameras on their own VLAN before tuning camera bitrates.
Recommended action - place the NVR on the same switch as the majority of cameras to keep streams local, cap camera bitrate, and replace the inter-floor links with managed switches supporting link aggregation. This is a small device purchase, not a recable, so it fits the stated constraint.