A Switch Architecture for Time-Triggered Transmission with Best-Effort Delivery

Zonghui Li, Wenlin Zhu, Kang G. Shin, Hai Wan, Xiaoyu Song, Dong Yang, Bo AiarXiv 2023 · DOI · arXiv

Abstract

Industrial real-time control systems demand high reliability and deterministic timing, typically achieved through Time-Triggered (TT) networking. However, the rigid scheduling of TT frames often results in a "latency gap" where data sits idle waiting for its allocated slot. This paper introduces the Synergistic Switch Architecture (SWA), which bridges TT reliability with Best-Effort (BE) speed. By cloning TT frames and allowing them to "race" through the network via high-priority BE paths, SWA significantly reduces end-to-end latency while maintaining deterministic guarantees.

Introduction: The Latency Challenge in Industrial Networks

In aerospace and automotive environments, timing is critical. Traditional TT networks like TTEthernet or Time-Sensitive Networking (TSN) schedule every packet to the nanosecond to prevent collisions. While this ensures delivery, it introduces inefficiency. A frame may be ready for transmission but must remain in a switch buffer until its precise schedule arrives, even if the network is idle. Conversely, standard Best-Effort (BE) Ethernet is fast but suffers from jitter and packet loss under congestion.

The "latency gap" can be substantial. In a typical fast Ethernet setup, a TT frame might be scheduled with a 200$\mu s$ delay for safety, despite a physical transit time of only 20$\mu s$. SWA aims to reclaim this 10x difference by treating BE delivery as an opportunistic shortcut.

The Synergistic Switch Architecture (SWA)

The core of SWA is a "cloning and racing" mechanism implemented within the switch hardware. When a TT frame enters a switch, the architecture creates a high-priority "cloned" copy. The original frame follows the strict TT schedule, while the clone is sent immediately via BE delivery.

The SWA processing pipeline, including classification, dual-path processing, and arrival filtering.

The SWA pipeline consists of five key stages:

  1. Classifier: Identifies incoming TT frames and generates clones.
  2. BE/TT Processing: The original frame is handled by the TT process (scheduled), while the clone enters the BE process (immediate).
  3. Sequence Checking: Uses strictly increasing sequence numbers to ensure that late clones from previous cycles do not interfere with current data.
  4. Arrival Filtering: At the destination or egress switch, the system accepts the first copy to arrive and discards the redundant one.
  5. Jitter Control: To maintain predictability, a "holding" strategy can be employed to ensure the final delivery falls within a pre-configured jitter window.

Experimental Evaluation

The researchers implemented SWA on Xilinx Virtex-7 FPGAs to validate its performance in a physical hardware environment. The testbed utilized a multi-switch topology to simulate industrial dataflow paths.

Physical topology and dataflow path used for evaluating the SWA architecture.

Evaluation used synthetic TT flows with varying frame lengths (128 to 512 bytes) and periods. To stress-test the system, an IXIA standard tester generated 100 Mbps of background "disturbance" BE traffic.

Key Results

  • Latency Reduction: For 128-byte frames, SWA reduced maximum end-to-end latency from 94.25$\mu s$ to approximately 64.34$\mu s$.
  • Congestion Resilience: Even under heavy background noise, the high-priority clones consistently arrived faster than scheduled TT frames.
  • Self-Recovery: Because cloning occurs at every switch, the system is resilient to individual packet loss. If a clone is dropped at one hop, the next switch generates a new clone from the scheduled TT frame it receives.
  • Jitter Management: The architecture successfully maintained jitter within a 10$\mu s$ window, meeting the requirements of sensitive industrial sensors.

Conclusion

The Synergistic Switch Architecture offers a robust solution for low-latency industrial networking. By doubling traffic for critical flows—a negligible cost given the low bandwidth of control traffic—SWA achieves significant speed gains without compromising the ironclad reliability of time-triggered schedules. Future work will explore safe-jitter calculations for larger, multi-hop topologies and further optimize the hardware complexity of the FPGA implementation.

Summary updated

Abstract

In Time-Triggered (TT) or time-sensitive networks, the transmission of a TT frame is required to be scheduled at a precise time instant for industrial distributed real-time control systems. Other (or {\em best-effort} (BE)) frames are forwarded in a BE manner. Under this scheduling strategy, the transmission of a TT frame must wait until its scheduled instant even if it could have been transmitted sooner. On the other hand, BE frames are transmitted whenever possible but may miss deadlines or may even be dropped due to congestion. As a result, TT transmission and BE delivery are incompatible with each other. To remedy this incompatibility, we propose a synergistic switch architecture (SWA) for TT transmission with BE delivery to dynamically improve the end-to-end (e2e) latency of TT frames by opportunistically exploiting BE delivery. Given a TT frame, the SWA generates and transmits a cloned copy with BE delivery. The first frame arriving at the receiver device is delivered with a configured jitter and the other copy ignored. So, the SWA achieves shorter latency and controllable jitter, the best of both worlds. We have implemented SWA using FPGAs in an industry-strength TT switches and used four test scenarios to demonstrate SWA's improvements of e2e latency and controllable jitter over the state-of-the-art TT transmission scheme.
BibTeX
@misc{li2023switch,
  author = {Zonghui Li and Wenlin Zhu and Kang G. Shin and Hai Wan and Xiaoyu Song and Dong Yang and Bo Ai},
  title = {{A Switch Architecture for Time-Triggered Transmission with Best-Effort Delivery}},
  year = {2023},
  eprint = {2309.11902},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2309.11902},
}
Figure