Quality-of-Service versus Realtime

That QoS and Realtime are mutually exclusive is shown by use of simple queuing theory. QoS mechanisms will not be active in the light to moderate (up to 50% average and 80% peak occupancy) communications load regime used by realtime systems, as by definition there is enough communications capacity to satisfy any and all requests in that moderate-load regime. QoS is required only when communications system path occupancy exceeds about 80%, where offered traffic must be forcibly shaped or deferred to fit the insufficient available communications capacity. No realtime system should be designed to operate in that regime, because the resulting large and random-appearing variations in latency make it impossible to achieve realtime behavior, or in many cases, to even achieve stable system operation.