Deadline Miss Ratio

Deadline miss ratio is the fraction of tasks in a real-time system that fail to complete before their deadlines, expressed as a dimensionless proportion: misses divided by total scheduled tasks. It is the most direct measure of a system's failure to meet its temporal obligations, and in hard real-time contexts — avionics, medical devices, automotive control — it is not a performance metric but a safety metric. A deadline miss ratio above zero means the system has failed its proof obligation.

The ratio is deceptively simple to calculate and notoriously difficult to control. A miss ratio of 0.01% sounds acceptable in a web server; in a pacemaker, it is a recall event. The metric also conceals information: it does not say *which* tasks missed, *why* they missed, or whether the misses were clustered in time. A system with a 1% miss ratio that concentrates all misses during system startup may be safer than a system with a 0.1% miss ratio that misses randomly during steady-state operation. The ratio is a scalar summary of a vector problem.

In soft real-time systems, the miss ratio is supplemented by other metrics: the average tardiness of late tasks, the maximum tardiness ever observed, and the jitter in completion times. But even here, the miss ratio dominates the conversation because it is the easiest to measure and the hardest to explain away. When a stakeholder asks 'does the system work?' the miss ratio is the number they look at first — and often the only number they understand.

The obsession with deadline miss ratio as a single figure of merit has produced systems that optimize for the metric rather than the mission. A scheduler that starves low-priority tasks to keep the miss ratio at zero is not a successful scheduler. It is a scheduler that has redefined 'success' to exclude the tasks it cannot save.