A Timing-Anomaly Free Dynamic Scheduling on Heterogeneous Systems

TL;DR

This work addresses timing anomalies in dynamic scheduling on heterogeneous multicore systems by introducing Deterministic Dynamic Execution (DDE), which imposes deterministic constraints on resource allocation and execution order to guarantee TA-free behavior. The authors develop a formal execution-progress model, define a strict TA, and prove that the proposed constraints ensure monotonic progress, making the all-WCET offline WCRT a safe bound. They provide two offline constraint-generation approaches—trace-based extraction and a HACPA heuristic—and implement DDE with a delayed-execution runtime to enforce the offline plan. Empirically, DDE eliminates timing anomalies and reduces WCRT and jitter across varied DAGs, with HACPA offering further improvements, while average AVRT changes are modest. Overall, this work delivers a rigorous TA-free scheduling framework for heterogeneous systems with practical impact on real-time performance guarantees.

Abstract

Heterogeneous systems commonly adopt dynamic scheduling algorithms to improve resource utilization and enhance scheduling flexibility. However, such flexibility may introduce timing anomalies, wherein locally reduced execution times can lead to an increase in the overall system execution time. This phenomenon significantly complicates the analysis of Worst-Case Response Time (WCRT), rendering conventional analysis either overly pessimistic or unsafe, and often necessitating exhaustive state-space exploration to ensure correctness. To address this challenge, this paper presents the first timing-anomaly-free dynamic scheduling algorithm for heterogeneous systems, referred to as Deterministic Dynamic Execution. It achieves a safe and tight WCRT estimate through a single offline simulation execution. The core idea is to apply deterministic execution constraints, which partially restrict the resource allocation and execution order of tasks at runtime. Based on a formally defined execution progress model for heterogeneous system scheduling, we prove the correctness of the proposed execution constraints and their ability to eliminate timing anomalies. Furthermore, we propose two methods to generate execution constraints. The first method derives execution constraints directly from the execution traces produced by existing scheduling algorithms. The second method is a heuristic-based approach that constructs execution constraints, enabling further reduction of the WCRT. Experimental results on synthetically generated DAG task sets under various system configurations demonstrate that, compared to traditional dynamic scheduling algorithms, our approach not only eliminates timing anomalies but also effectively reduces both the WCRT and response time jitter.

Figures (12)

• Figure 1: An example of a multi-typed DAG with seven nodes (tasks) and corresponding processing units with execution times for each node.
• Figure 2: Two schedules of the DAG in Fig. 1 under HBFS: (a). an arbitrary online execution with response time 15, and (b). all-WCETs offline simulated execution with response time 18, which is always considered as WCRT.
• Figure 3: An example of a TA for the DAG shown in Fig. \ref{['fig:DAG1']} under the HBFS scheduling, where the execution time of the tasks highlighted in green is shorter than the WCET in Fig. \ref{['fig:twoExectionCase']}(b), resulting in a response time of 20.5. It exceeds the WCRT of 18 obtained from the all-WCETs execution in Fig. \ref{['fig:twoExectionCase']}(b).
• Figure 4: A conservative simulated execution by transforming the faster processing units in Fig. \ref{['fig:DAG1']} to the slowest type within the same architecture and performing an all-WCETs estimation, the resulting WCRT may still not be safe. The obtained WCRT is 20 remains smaller than the 20.5 in Fig. \ref{['fig:TA-case']}.
• Figure 5: System execution progress model and partial order of progresses

Theorems & Definitions (5)

• proof
• proof
• proof
• proof
• proof