Hardware/Algorithm Co-design for Real-Time I/O Control with Improved Timing Accuracy and Robustness

Zhe Jiang, Shuai Zhao, Ran Wei, Xin Si, Gang Chen, Nan Guan

TL;DR

This paper introduces ROTA-I/O, a novel hardware-software co-designed system that significantly improves the timing accuracy and robustness of I/O operations in safety-critical real-time systems. By employing hardware-based Execution Time Servers (ETSs) and a two-level scheduler, ROTA-I/O effectively isolates tasks and mitigates the impact of unexpected timing defects, outperforming existing solutions without substantial overhead.

What Problem Does This Paper Address?

In safety-critical systems, such as autonomous vehicles or spacecraft, precise timing of I/O operations (interactions with sensors and actuators) is paramount. Inaccurate timing can lead to incorrect control, faulty environmental readings, and potentially catastrophic system failures. While dedicated hardware solutions have been proposed to address this, they often prove fragile when faced with "timing defects" – unforeseen issues like incorrect execution time calculations or transient hardware failures. Existing software approaches struggle to manage the increasing complexity of modern hardware and the inherent uncertainties in I/O transmissions. This paper tackles the challenge of achieving both high timing accuracy and robust performance in I/O control, even in the presence of these unpredictable timing defects.

Methods

The core innovation lies in the ROTA-I/O co-processor, which integrates hardware and algorithms for I/O management. Key components include:

  • Execution Time Servers (ETSs): These hardware modules provide temporal isolation between I/O tasks. By limiting the impact of timing defects within one ETS, they prevent cascading failures across the system. Each ETS is allocated a time budget and a start time.
  • Two-Level Scheduler: A hierarchical scheduler manages ETSs. A global scheduler allocates time budgets to ETSs, while local schedulers within each ETS prioritize tasks. This structure allows for configurable scheduling infrastructure.
  • ROTA-Sched Algorithm: This ETS-based scheduling algorithm configures the ETS parameters (start time, budget) to ensure timing predictability and robustness. It first identifies "exact-accurate" jobs that must meet precise deadlines and assigns them dedicated ETSs. Remaining "quality-based" jobs are then scheduled to maximize their quality within available time slots, potentially sharing ETSs. Crucially, ROTA-Sched configures ETSs with an "additional capacity" to absorb unexpected timing delays, preventing deadline misses in subsequent tasks.
  • Dedicated ISA: A custom Instruction Set Architecture (ISA) is defined for configuring ROTA-I/O, enabling efficient communication between the processor and the co-processor.

Datasets

Not discussed in the paper. The evaluation uses synthetic workloads and automotive safety tasks.

Key Insights and Ideas

  • Hardware-Software Co-design is Crucial: The paper strongly advocates for a combined hardware and software approach, demonstrating that neither discipline alone can fully address the complex timing and robustness requirements of modern real-time I/O.
  • Temporal Isolation is Key to Robustness: The ETS mechanism is a powerful concept for preventing timing defects from propagating and causing system-wide failures. This is a significant departure from purely software-based solutions that struggle with such isolation.
  • Proactive Delay Absorption: The concept of configuring ETSs with an "additional capacity" to absorb timing defects before they cause deadline misses is a novel and effective strategy for enhancing robustness. This is a more proactive approach than simply reacting to failures.
  • Scalability without Critical Paths: The micro-architecture of ROTA-I/O is designed to scale with the number of cores and I/Os without introducing significant performance bottlenecks, which is a critical requirement for modern many-core systems.

Key Results

  • Improved Timing Accuracy and Robustness: ROTA-I/O demonstrated a significant improvement in acceptance ratio (percentage of jobs meeting deadlines) and normalized operation quality compared to state-of-the-art (SOTA) methods, especially under increasing timing defect severity. For instance, it achieved 22.61% higher acceptance ratio and 42.32% higher operation quality on average when timing defects were significant.
  • Low Hardware Overhead: The ROTA-I/O co-processor incurred a negligible hardware overhead (less than 3% of a full-featured processor) and was comparable to standard Ethernet controllers in terms of LUTs and registers.
  • Reduced Software Overhead: By offloading I/O management to hardware, ROTA-I/O reduced software overhead by 32 KB compared to a software-only solution.
  • Scalability: The system demonstrated good scalability, with area consumption remaining nearly constant as the number of cores increased, and linearly increasing with the number of I/Os. Maximum frequency was consistently higher than baseline SoCs.
  • System-Level Performance: In an automotive case study, ROTA-I/O consistently outperformed baseline systems in terms of system-level real-time performance (Success Ratio) and I/O quality, even with injected hardware faults.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that while ETSs limit propagation between ETSs, jobs within an ETS could still miss deadlines or be terminated if the ETS capacity is exhausted by timing defects.
  • The scalability analysis suggests that for very complex many-core systems, considering Network-on-Chip (NoC) contention and interference might be necessary for further optimization.
  • Future Work:
    • The authors plan to tape out ROTA-I/O on safety-critical micro-controllers in a real chip to evaluate its energy efficiency under different application scenarios.
  • Further investigation into mitigating NoC contention in highly integrated many-core SoCs is suggested.

Summary updated

Abstract

In safety-critical systems, timing accuracy is the key to achieving precise I/O control. To meet such strict timing requirements, dedicated hardware assistance has recently been investigated and developed. However, these solutions are often fragile, due to unforeseen timing defects. In this paper, we propose a robust and timing-accurate I/O co-processor, which manages I/O tasks using Execution Time Servers (ETSs) and a two-level scheduler. The ETSs limit the impact of timing defects between tasks, and the scheduler prioritises ETSs based on their importance, offering a robust and configurable scheduling infrastructure. Based on the hardware design, we present an ETS-based timing-accurate I/O schedule, with the ETS parameters configured to further enhance robustness against timing defects. Experiments show the proposed I/O control method outperforms the state-of-the-art method in terms of timing accuracy and robustness without introducing significant overhead.
BibTeX
@inproceedings{jiang2024hardware,
  author = {Zhe Jiang and Shuai Zhao and Ran Wei and Xin Si and Gang Chen and Nan Guan},
  title = {{Hardware/Algorithm Co-design for Real-Time I/O Control with Improved Timing Accuracy and Robustness}},
  booktitle = {RTSS},
  year = {2024},
  doi = {10.1109/rtss62706.2024.00024},
}
Figure