Schedulability analysis and stack size minimization for adaptive mixed criticality scheduling with semi-Clairvoyance and preemption thresholds

https://doi.org/10.1016/j.sysarc.2021.102383Get rights and content

Abstract

In cost-sensitive Mixed-Criticality Systems (MCS) such as automotive Electrical/Electronic (E/E) systems, it is important to reduce application memory footprint, since such a reduction may enable significant cost savings due to adoption of cheaper processor models. Preemption Threshold Scheduling (PTS) is an effective technique for controlling the degree of preemption and reducing system stack usage of a multitasking system under priority-based scheduling. In this paper, we present a schedulability test to enable integration of Preemption Threshold Scheduling (PTS) and Adaptive Mixed Criticality (AMC) scheduling with Semi-Clairvoyance, which can be used by high-level optimization algorithms with different possible optimization objectives, including improving schedulability and reducing system stack usage of an MCS. The experimental results indicate that our approach results in significant benefits in terms of improving schedulability and reducing system stack usage compared to state-of-the-art.

Section snippets

Introduction and related work

Certain application domains, e.g., Automotive Electrical/Electronic (E/E) systems, are very resource-constrained due to size, weight and power (SWaP) constraints, as well as monetary cost constraints thanks to intense cost-cutting pressures in today’s competitive consumer marketplace. Auto-makers are very cost-sensitive due to the mass production of vehicles. As a modern high-end vehicle may contain tens to 100+ Electronic Control Units (ECUs) in it, and each vehicle model may sell millions of

PT-AMC-sem and its schedulability test

We consider a mixed-criticality task set with N independent sporadic real-time tasks Γ={τi|1iN} on a uniprocessor. Based on Vestal’s task model [13], each task τi is given multiple Worst-Case Execution Time (WCET) estimates, one for each criticality level. We consider a dual-criticality MCS with two criticality levels: Low (LO) and High (HI). Task τi has a tight, optimistic WCET estimate CiLO in Low-Criticality (abbreviated as LO-crit) mode that may be occasionally exceeded, and a loose,

Illustrating example

We use an example to illustrate the schedulability tests. Consider a task set with three tasks, with the parameters shown in Table 1. To apply the analysis of PT-AMC-sem, we have assigned priorities and PTs to tasks in the running example. Let us consider HI-crit task τ3 as the task under analysis. According to Eq. (1), it is schedulable if R3LOD3max(R3LO,R3HI,R3CC)D3, where D3=50.

First, let us compute τ3’s WCRT in LO-crit mode. We have hep(τ3)={τ1,τ2,τ3}, ht(τ3)={τ1}, hp(τ3)={τ1,τ2}, hpL(τ

Performance evaluation

We use the same task set generation methodology as [24] to generate synthetic task sets for our experiments, with the following parameters:

  • Each task τi’s utilization in LO-crit mode(UiLO=CiLO/Ti) is generated using UUnifast [28], providing an unbiased distribution, with default task set utilization in LO-crit mode set to 0.85.

  • Each task τi’s period Ti is generated according to a log-uniform distribution [29] in the interval [10,200] ms.

  • Each task τi’s deadline Di is set to be equal to its period T

Conclusions and future work

In this paper, we present a schedulability test to enable integration of Preemption Threshold Scheduling (PTS) with Mixed-Criticality Scheduling with Semi-Clairvoyance. Performance evaluation indicates that our approach results in significant benefits in terms of improving schedulability and reducing system stack usage compared to Fully Preemptive Scheduling, which helps to implement MCS more efficiently in resource-constrained embedded systems. As part of future work, we plan to consider

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

This work was supported by the National Natural Science Foundation of China (Grant No. 61902185 and 61877015), the Jiangsu Provincial Natural Science Foundation (Grant No. BK20190448 and BK20190447), and “111” Program (No. B13022).

Qingling Zhao received the Ph.D. degree from Zhejiang University(2015). She is currently an associate professor in the school of computer science and engineering at the Nanjing University of Science and Technology. Her research interests include real-time systems, mixed-criticality systems, and cyber-physical systems.

References (31)

  • WangC. et al.

    Global fixed priority scheduling with preemption threshold: Schedulability analysis and stack size minimization

    IEEE Trans. Parallel Distrib. Syst.

    (2016)
  • WangC. et al.

    Minimizing stack memory for hard real-time applications on multicore platforms with partitioned fixed-priority or EDF scheduling

    ACM Trans. Design Automa. Electron. Syst. (TODAES)

    (2016)
  • GuZ. et al.

    Cache-partitioned preemption threshold scheduling

    ACM Trans. Embed. Comput. Syst.

    (2016)
  • BrilR.J. et al.

    Fixed priority scheduling with pre-emption thresholds and cache-related pre-emption delays: integrated analysis and evaluation

    Real-Time Syst.

    (2017)
  • VestalS.

    Preemptive scheduling of multi-criticality systems with varying degrees of execution time assurance

  • Cited by (5)

    • Towards an energy-efficient quarter-clairvoyant mixed-criticality system

      2022, Journal of Systems Architecture
      Citation Excerpt :

      These frameworks, which trigger a mode switch when they detect over-execution of a Hi-task, are also called “non-clairvoyant MCSs” [12,13]. Different from non-clairvoyant MCSs, Baruah et al. [12] and Agrawal et al. [13] introduce clairvoyant and semi-clairvoyant MCS theoretical models, which assume that whether a Hi-task will overrun its Lo-WCET is known before or at release [14,15]. As shown in [12,13], both clairvoyant and semi-clairvoyant MCSs outperform non-clairvoyant MCSs.

    • Criticality-Aware EDF Scheduling for Constrained-Deadline Imprecise Mixed-Criticality Systems

      2024, IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
    • Compensating Adaptive Mixed Criticality Scheduling

      2022, ACM International Conference Proceeding Series

    Qingling Zhao received the Ph.D. degree from Zhejiang University(2015). She is currently an associate professor in the school of computer science and engineering at the Nanjing University of Science and Technology. Her research interests include real-time systems, mixed-criticality systems, and cyber-physical systems.

    Mengfei Qu He is currently a master graduate student in the school of computer science and engineering at the Nanjing University of Science and Technology. His research interests include mixed criticality system, embedded system, etc.

    Bo Huang received the Ph.D. degree from Nanjing University of Science and Technology (2006). He is currently a professor in the school of computer science and engineering at Nanjing University of Science and Technology. His research interests include discrete event systems, Petri nets, system modeling and intelligent scheduling.

    Zhe Jiang received his Ph.D. from University of York (2019). He is currently working as the system design engineer in ARM Ltd and visit research associate in University of York. He is research int rests include safety-critical system, system architecture, and system micro-architecture. He can be reached at: [email protected].

    Haibo Zeng (Member, IEEE) is with Department of Electrical and Computer Engineering at Virginia Tech. He received his Ph.D. in Electrical Engineering and Computer Sciences from University of California at Berkeley. He was a Senior Researcher at General Motors R&D until October 2011, and an Assistant Professor at McGill University until August 2014. His research interests are embedded systems, cyber-physical systems, and real-time systems. He received five best / outstanding paper awards in the above fields.

    1

    Co-first authors of the article.

    View full text