The Trade Show That Breaks the Network Twice a Year
A 480-room convention hotel hosts a 3-day trade show twice a year with roughly 2,200 concurrent devices across 30,000 sq ft of meeting space, served by 40 access points. During the last show, guest Wi-Fi collapsed each afternoon, POS terminals in three outlets timed out, and the PMS slowed at check-in peak. The hotel has a 200 Mbps circuit that monitoring shows averaging 55 Mbps except during the show. The owner's preferred fix is to double bandwidth. A vendor is proposing a full Wi-Fi 6E replacement with a three-year support contract. The hotel has no per-SSID shaping, one flat internal VLAN, and a DHCP scope sized for 800 leases.
1. What would you measure before approving any spend, and what would each measurement rule in or out?
Read the worked answer
The symptoms split across three layers, and the fix for one does nothing for the others. Instrument for a week, then replay the last show's afternoon profile.
- Circuit utilisation at 15-minute granularity during the failure window. A 55 Mbps average with a 200 Mbps circuit suggests the WAN is not the bottleneck. If peak utilisation never approaches the committed rate but latency and loss rise, the problem is upstream of bandwidth.
- Airtime utilisation and per-AP client counts. 40 APs for 2,200 devices is roughly 55 clients per radio. If the failure correlates with high airtime utilisation and retransmits rather than with WAN throughput, adding bandwidth is money spent on the wrong layer.
- DHCP pool exhaustion. A scope sized for 800 leases against 2,200 devices will starve new associations and look exactly like a Wi-Fi failure. This is the cheapest thing on the list to test and the most embarrassing to miss.
- POS timeout correlation. If POS timeouts track the same minutes as Wi-Fi collapse, they share a cause (airtime or upstream congestion). If they are independent, look for duplex mismatch, spanning-tree or an undersized uplink on the switch serving those outlets.
- Channel plan and low-rate support. Overlapping channels on 2.4 GHz and legacy low data rates consume airtime disproportionately.
Each measurement maps to a different purchase, which is the point: without them the decision defaults to whoever has a proposal ready.
2. Evaluate the 'just double the bandwidth' proposal and set out a remediation sequence with funding logic.
Read the worked answer
The proposal. Doubling to 400 Mbps would raise committed cost permanently to fix a fault that occurs roughly six days a year, and it cannot fix DHCP exhaustion, airtime contention or a flat VLAN. It may not even be available on the building's fibre. Treat it as a possible component, not the plan.
Sequence.
- Zero-cost fixes now: expand the DHCP scope, disable low data rates, fix the channel plan, and segment into separate SSIDs with VLANs for guest, POS/corporate, and building-management/IoT. Segmentation is also a PCI scope reduction, so it earns its cost twice.
- Configuration and policy: per-SSID rate limits and QoS so one guest streaming video cannot starve POS traffic. This is often the difference between 'collapsed' and 'degraded'.
- Capacity: add APs where density is worst - 30,000 sq ft of meeting space with 2,200 devices is the real problem - and only then re-size the circuit against measured concurrency, with a burstable option rather than a flat increase.
- Commercial: price show-week bandwidth into the meeting-space rate card so peak capacity is funded by the events that cause it.
Funding logic. Steps 1–2 are opex and internal labour, deliverable inside 30 days, and buy credibility. Step 3 is capex; the Wi-Fi 6E vendor is only credible after steps 1–2, because otherwise the symptom returns on new hardware and the three-year contract is already signed.