High-Density Auditorium: Coverage Passes, Polling Fails
Westbrook University's 1,200-seat auditorium hosts first-year lectures with mandatory live polling. Average attendance is 1,150 students, each arriving with a laptop and a phone, about 2,300 devices. Eight access points sit in the ceiling grid, configured for 40 MHz on 5 GHz, 2.4 GHz at full power, and no band steering. At lecture start the controller reports 1,900 associated clients, 62% channel utilization on the least-used 5 GHz channel, a median application round-trip of 340 ms, and 9% of poll submissions failing. The client mix is 60% Wi-Fi 5, 25% Wi-Fi 6 or 6E, and 15% older devices. The university has approved a budget for new access points but will not fund new cable pulls; the room has 12 existing ceiling data drops, all Cat6, each with a 1 GbE port. One Saturday maintenance window is available before term starts, with a 200-device pilot permitted two weeks earlier.
1. Is this a coverage problem or a capacity problem? Support the conclusion with the numbers and identify what actually limits the room.
Read the worked answer
It is a capacity and airtime problem, not coverage. 2,300 devices over eight APs is about 287 devices per AP, or roughly 144 per radio, against a practical high-density design target of 50–75 active clients per radio for interactive applications. The 40 MHz channels make it worse: 40 MHz halves the reuse plan to about four usable channels in UNII-1 and UNII-3, so eight APs in one room means two APs per channel at close range, which is exactly the 62% utilization and 340 ms round-trip observed. Polling is low-bandwidth but high-concurrency: every response is a small frame competing for medium access, and association and management traffic for 1,900 clients consumes airtime that the poll cannot use. 2.4 GHz at full power also pulls devices onto a three-channel band. The symptom set, healthy RSSI with poor delivery, is the signature of an airtime shortage rather than coverage. Adding power changes nothing; adding radios and channels does.
2. Given twelve cable drops and no funding for more, specify the design you would deploy and justify the trade-offs.
Read the worked answer
Twelve drops means twelve tri-band APs, so 36 radios instead of 16. Set every radio to 20 MHz, which triples the 5 GHz reuse plan and lets adjacent APs sit on different channels; the throughput lost per client is more than recovered by reduced contention. Mount APs in the ceiling grid clear of the stage and the front rows, aimed into the seating, and keep transmit power modest so cells stay small. 6 GHz serves only the 25% of devices that are 6E, and low-power-indoor limits shrink those cells further, so treat 6 GHz as extra capacity and keep 5 GHz as the anchor for the Wi-Fi 5 majority. Reduce SSIDs to two, disable 2.4 GHz low rates, enable 802.11k/v for steering, and set a minimum basic rate so management frames do not burn the lowest rates. If the presenter or lecture-capture devices can be wired, do it: removing 40 devices from the ceiling is free capacity.
3. Design the pilot and the go-live validation. What would you measure, and what acceptance criteria would you set before the Saturday window?
Read the worked answer
Run the pilot at 200 devices in the same seating geometry, not in an empty room, and measure outcomes rather than signal. Capture association success rate, time to associate under load, poll submission success rate at the application layer, median and 95th-percentile application round-trip, per-radio client counts, retry rate, and channel utilization from the AP counters. Useful targets: at least 99% association success, 95% of clients associated within 10 seconds of the lecture starting, poll success at 99% or better, and utilization headroom remaining under 50% at the pilot load. Then extrapolate: a 200-device pilot passing does not guarantee 2,300 devices pass, so the Saturday window must be followed by a full-scale rehearsal with volunteers before term, or the first lecture becomes the test. If the pilot degrades, the fallback is a wired dock for polling or restricting video, agreed with the faculty in advance rather than discovered live.