A Survey of Real-time Scheduling on Accelerator-based Heterogeneous Architecture for Time Critical Applications

TL;DR

This survey addresses timing-critical computing on accelerator-based heterogeneous architectures (CPU+PEs such as GPUs/TPUs/FPGA) by integrating architecture, task models, and real-time scheduling approaches for time-sensitive applications. It introduces task models including the Self-Suspension Segmented Model with WCETs $C_i^m$ and $A_i^m$, DAGs $G_i=(V_i,E_i)$, and processing chains $\Gamma_c$, and reviews soft and hard real-time strategies across vendor/Linux baselines, single- and multi-accelerator designs, and application-driven scenarios. Key contributions include a comprehensive taxonomy of architectures and task models, analysis of memory-copy and data-movement overheads, and discourse on spatial/temporal partitioning, preemption, and MILP-based allocations, complemented by energy/thermal-aware hard RT approaches and diverse application case studies such as autonomous systems, perception, language models, satellites, and XR. The paper also highlights open challenges—schedulability blow-up with more processor types, non-preemptive data transfers, and the lack of standardized evaluations—and calls for cohesive benchmarks and methodology to enable rigorous, fair comparisons across heterogeneous RT scheduling solutions.

Abstract

Accelerator-based heterogeneous architectures, such as CPU-GPU, CPU-TPU, and CPU-FPGA systems, are widely adopted to support the popular artificial intelligence (AI) algorithms that demand intensive computation. When deployed in real-time applications, such as robotics and autonomous vehicles, these architectures must meet stringent timing constraints. To summarize these achievements, this article presents a comprehensive survey of real-time scheduling techniques for accelerator-based heterogeneous platforms. It highlights key advancements from the past ten years, showcasing how proposed solutions have evolved to address the distinct challenges and requirements of these systems. This survey begins with an overview of the hardware characteristics and common task execution models used in accelerator-based heterogeneous systems. It then categorizes the reviewed works based on soft and hard deadline constraints. For soft real-time approaches, we cover real-time scheduling methods supported by hardware vendors and strategies focusing on timing-critical scheduling, energy efficiency, and thermal-aware scheduling. For hard real-time approaches, we first examine support from processor vendors. We then discuss scheduling techniques that guarantee hard deadlines (with strict response time analysis). After reviewing general soft and hard real-time scheduling methods, we explore application- or scenario-driven real-time scheduling techniques for accelerator-enabled heterogeneous computing platforms. Finally, the article concludes with a discussion of open issues and challenges within this research area.

Figures (6)

• Figure 1: Real-time scheduling of parallel tasks on the heterogeneous architecture.
• Figure 2: Overview of this survey.
• Figure 3: Different task modeling approaches.
• Figure 4: Tree Diagram for Soft Real-Time Tasks Scheduling.
• Figure 5: Tree Diagram for Hard Real-Time Tasks Scheduling.