[Paper Review] Parametric Schedulability Analysis of Fixed Priority Real-Time Distributed Systems

Youcheng Sun, Romain Soulat+3 more|arXiv (Cornell University)|Feb 6, 2013

Real-Time Systems Scheduling15 references4 citations

TL;DR

This paper proposes two novel approaches for parametric schedulability analysis in fixed-priority real-time distributed systems: an analytic method extending classical sensitivity analysis to distributed, preemptive, and non-preemptive scheduling with jitters and unconstrained deadlines, and a model checking approach using Parametric Timed Automata (PTA) with the Inverse Method. The analytic method achieves performance close to holistic analysis, while the PTA approach ensures completeness but is slower, offering a trade-off between efficiency and coverage across parameter spaces.

ABSTRACT

Parametric analysis is a powerful tool for designing modern embedded systems, because it permits to explore the space of design parameters, and to check the robustness of the system with respect to variations of some uncontrollable variable. In this paper, we address the problem of parametric schedulability analysis of distributed real-time systems scheduled by fixed priority. In particular, we propose two different approaches to parametric analysis: the first one is a novel technique based on classical schedulability analysis, whereas the second approach is based on model checking of Parametric Timed Automata (PTA). The proposed analytic method extends existing sensitivity analysis for single processors to the case of a distributed system, supporting preemptive and non-preemptive scheduling, jitters and unconstrained deadlines. Parametric Timed Automata are used to model all possible behaviours of a distributed system, and therefore it is a necessary and sufficient analysis. Both techniques have been implemented in two software tools, and they have been compared with classical holistic analysis on two meaningful test cases. The results show that the analytic method provides results similar to classical holistic analysis in a very efficient way, whereas the PTA approach is slower but covers the entire space of solutions.

Motivation & Objective

• To address the challenge of designing robust real-time distributed systems where small parameter variations (e.g., task execution times) can compromise schedulability.
• To extend single-processor parametric schedulability analysis to distributed systems with fixed-priority scheduling, supporting both preemptive and non-preemptive execution.
• To provide a comprehensive analysis of the feasible parameter space (e.g., computation times, deadlines) ensuring all tasks meet their end-to-end deadlines.
• To compare two distinct parametric analysis techniques—analytic extension of classical schedulability tests and PTA-based model checking—on accuracy, efficiency, and scalability.

Proposed method

• The analytic method extends Bini et al.'s sensitivity analysis to distributed systems, supporting computation time, deadline, and jitter parameters with both preemptive and non-preemptive scheduling.
• It uses a modified version of the holistic schedulability test to compute feasible regions in the parameter space, enabling efficient exploration of design trade-offs.
• The PTA-based approach models the entire distributed system as a Parametric Timed Automaton, capturing all possible execution behaviors including timing constraints and communication delays.
• Model checking is performed using the Inverse Method to compute the exact schedulability region in the parameter space, ensuring completeness and correctness.
• Both methods are implemented in dedicated tools: RTSCAN for the analytic approach and IMITATOR for the PTA-based method.
• Classical holistic analysis via the MAST tool is used as a baseline for comparison in terms of precision and execution time.

Experimental results

Research questions

• RQ1 Can classical schedulability analysis be extended to support parametric analysis in distributed real-time systems with fixed priority scheduling?
• RQ2 How does the performance and accuracy of the analytic parametric method compare to classical holistic analysis in terms of execution time and coverage?
• RQ3 To what extent can PTA-based model checking provide a complete and precise characterization of the schedulability region in distributed systems?
• RQ4 What are the trade-offs between efficiency (execution time) and completeness (coverage of the feasible region) between the analytic and PTA-based approaches?

Key findings

• The analytic method (RTSCAN) achieves performance close to classical holistic analysis, with execution times of 0.27s for test case 1 and 1m47s for test case 2b, significantly outperforming PTA-based methods.
• MAST, using binary search for slack computation, required 7.19s for test case 1 and up to 40 minutes for test case 2a, showing scalability issues with larger hyperperiods.
• IMITATOR, while ensuring completeness, was significantly slower—taking 19 minutes 42 seconds for test case 1 and over 4 hours for test case 2a—due to repeated region computation and inefficient neighborhood search.
• The PTA approach identified 257 schedulability regions in test case 2a, indicating high computational cost from fragmented region exploration.
• The analytic method demonstrated high efficiency and accuracy comparable to holistic analysis, making it suitable for incremental design and parameter space exploration.
• The authors identify that smarter parameter space exploration strategies, such as using the analytic method to initialize PTA exploration, could drastically improve IMITATOR’s performance.