Real-Time Computing
Real-time computing guarantees not just correct answers but answers delivered within a strict, predictable time bound.
Correct and on time
In real-time computing, a result that is correct but late is a failure. What matters is meeting a deadline every time. This is a different requirement from ordinary computing, where average speed matters more than worst-case timing.
Hard versus soft
- Hard real-time: missing a deadline is a system failure, as in machine protection.
- Soft real-time: occasional misses degrade quality but are tolerable, as in a display update.
Determinism over speed
Real-time systems prize predictability. A fast average with an occasional long pause — from memory management, caching, or scheduling — can be worse than a slightly slower system that never stalls. Design focuses on bounding the worst case, not just improving the typical case.
Where it applies
Plasma control loops, magnet protection, and interlocks must respond within fixed time windows measured in milliseconds or less. When a fault condition appears, the response cannot wait on a busy general-purpose system; it runs on infrastructure engineered to guarantee its timing.
Layered timing
A plant separates fast, hard real-time protection and control from slower advisory computation. Machine learning and heavy analysis run in the slower layer, feeding the fast layer with prepared results, so the guarantees of the real-time core are never at the mercy of an unpredictable model.