Analysis of Fault Tolerant Multi-stage Switch Architecture for TSN

TL;DR

This paper evaluates a multi-stage switch architecture for Time-Sensitive Networking (TSN), demonstrating improved fault tolerance but increased complexity and cost, with potential for safety-critical applications.

Contribution

It provides a feasibility analysis of multi-stage TSN switches, highlighting their advantages in fault tolerance and identifying trade-offs in performance and cost.

Findings

Multi-stage architecture offers better fault tolerance than single-stage.

Latency and throughput are comparable between architectures.

Multi-stage architecture has higher cost and complexity.

Abstract

We conducted the feasibility analysis of utilizing a highly available multi-stage architecture for TSN switches used for sending high priority, mission-critical traffic within a bounded latency instead of traditional single-stage architectures. To verify the TSN functionality, we implemented the 'strict priority' feature. We evaluated the performance of both architectures on multiple parameters such as fault tolerance, packet latency, throughput, reliability, path length effectiveness, and cost per unit. The fault tolerance analysis demonstrated that the multi-stage architecture fairs better than the single-stage counterpart. The average latency and throughput performance of multi-stage architectures, although low, can be considered comparable with single-stage counterparts. However, the multi-stage architecture fails to meet the performance of single-stage architectures on parameters such as reliability, path length effectiveness, and cost-effectiveness. The improved fault tolerance comes at the cost of increased hardware resources, cost, and complexity. However, with the advent of cost-effective technologies in hardware design and efficient architecture designs, the multi-stage switching architecture-based TSN switches can be made reasonably comparable to single-stage switching TSN switches. This work gives initial confidence that the multi-stage architecture can be pursued further for safety-critical systems that require determinism and reliability in the communication of critical messages.