Semi-Clairvoyance in Mixed-Criticality Scheduling

Abstract

In the Vestal model of mixed-criticality systems, jobs are characterized by multiple different estimates of their actual, but unknown, worst-case execution time (WCET) parameters. Prior work on mixed-criticality scheduling theory assumes that the execution duration of a job is only revealed by actually executing the job through to completion. We consider a different *semi-clairvoyant* model here, in which it is assumed that upon arrival a job reveals which of its WCET parameters it will respect. We identify circumstances under which this is a reasonable model, and design and evaluate scheduling algorithms appropriate for this model. We show that such semi-clairvoyance yields a significant quantifiable benefit over non-clairvoyance, in terms of both the complexity of schedulability analysis and the speedup needed to ensure schedulability.