A university campus supports 8,000 endpoints with voice, video and bulk data across many switches, and IT staff want scalable classification without per-flow signaling or reserved bandwidth on every hop. A small telepresence suite needs guaranteed admission in one building. Which QoS model choice best fits the campus-wide scale requirement?

DiffServ fits the campus-wide requirement because it marks traffic into aggregates with DSCP values and applies per-hop behaviors at each switch, providing differentiated treatment for voice, video and bulk data without maintaining per-flow state or signaling every hop, which scales to 8000 endpoints. Strict priority queuing everywhere is operationally common for latency reduction but fails as a model choice because without aggregate classification and PHB policy it cannot scalably differentiate classes and risks starving bulk data during congestion. Overprovisioned best-effort is widely used on modern campuses and avoids signaling overhead, but it fails the stated need for scalable classification because it provides no differentiation when voice and video contend with bursts. Building-only IntServ with RSVP is sound for the telepresence suite because limited-scope reservations enable true admission control and mathematical guarantees, but it fails the campus-wide scale prompt because extending reservations to thousands of endpoints reintroduces state and signaling overhead the stem rules out. Exam caveat: distinguish scope wording, campus-wide implies aggregates while single-suite guarantee implies admission control, and do not select a correct local solution when the question asks for global scale. Practical check: verify DSCP trust boundaries, queuing maps and EF/AF allocations on distribution switches, then confirm telepresence admission stays confined to its building with policers showing intended PHB markings under load.