On Bounding Execution Demand under Mixed-Criticality EDF

Abstract

This paper is concerned with mixed-criticality systems where a set of low-criticality (LO) and high-criticality (HI) tasks share one processor and are scheduled under the EDF algorithm. Basically, the system operates in two modes: LO and HI mode. In LO mode, one HI task may exceed its execution budget, which then causes a change to HI mode in the system. HI tasks are assigned larger execution budgets in HI mode at the cost of the LO tasks --- which are assumed to be discarded. Since these mode changes may happen at arbitrary points in time, it is difficult to find an accurate bound on the amount of carry-over execution demand. That is the execution demand by HI jobs that were released before, but did not finish executing at the point of changing to HI mode. As a consequence, the resulting characterization of the overall execution demand becomes pessimistic. In this paper, to overcome this problem, a technique is proposed that works around the computation of carry-over execution demand and results in a more accurate bound on execution demand under mixed-criticality EDF. In principle, the proposed technique consists in separating the schedulability analysis of stable HI mode from that of the transition between modes and deriving a separate demand bound function for the latter case. The proposed technique results not only in a considerably simpler, but also tighter bound on execution demand under mixed-criticality EDF, in particular, as the number of HI tasks increases. We illustrate this analytically and by a large set of experiments based on synthetic data.