IT lexicon › Infra & Ops › Mixed criticality

Mixed criticality

Infra & Ops På svenska → Updated: 2026-11-04

Running functions with different safety requirements — brakes and the car radio, say — on the same hardware, with the guarantee that one never suffers for the other.

Each function used to get its own control unit, until a modern car carried more than a hundred of them. Consolidating saves wiring, weight and money but makes certification hard: airbag code is assessed to ASIL D under ISO 26262, the media player barely at all. Either everything is certified to the highest level, which is ruinously expensive, or you prove the two cannot affect each other.

The research field is usually dated to Steve Vestal's 2007 paper at Honeywell, which proposed scheduling in which each task carries several WCET estimates — an optimistic one and a paranoid one. If a critical task overruns its optimistic budget, the less important work gets dropped. In practice the problem is mostly solved with partitioning via a hypervisor or separation kernel.

← Back to the lexicon