Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Cost-efficient Credit-Based Shaper Deployment Framework for Time-Sensitive Networks

Santiago Torres-Borda, Ahlem Mifdaoui

TL;DR

The paper tackles the cost barrier of full CBS deployment in Time-Sensitive Networking by proposing a partial CBS deployment framework that leverages Network Calculus for schedulability verification and a cost-aware heuristic to place and configure IdleSlopes on a minimal set of legacy devices. It formalizes the CBS deployment problem as an optimization (with ILP and linearized versions) and provides a scalable heuristic to cluster CBS while preserving end-to-end deadlines, validated on automotive TSN use cases. Key contributions include the NC-based schedulability framework, the local-deadline and minimum IdleSlope formulations, and a complexity-aware heuristic with demonstrated hardware cost reductions (up to 70% fewer TSN devices and 91% fewer CBS instances) and notable delay improvements for both shaped and non-shaped traffic. The results indicate practical impact for industrial networks by enabling deterministic QoS with substantially lower capital expenditure and easier integration into existing Ethernet infrastructures. Future work will extend evaluation to larger networks, analyze load variation and interference, and explore integration with framing techniques like Frame Replication and Elimination.

Abstract

Time-sensitive networks are designed to meet stringent Quality of Service (QoS) requirements for mixed-criticality traffic with diverse performance demands. Ensuring deterministic guarantees for such traffic while reducing deployment costs remains a significant challenge. This paper proposes a cost-efficient partial deployment strategy for Time Sensitive Networking (TSN) devices within legacy Ethernet network. At the core of our approach is the Credit-Based Shaper (CBS), a key TSN scheduling mechanism. Unlike cost-prohibitive full CBS deployment, our approach selectively integrates CBS where it is most needed to enhance performance while reducing costs. Combining Network Calculus for schedulability verification and a heuristic optimization method for CBS configuration and placement, our proposal minimizes deployment costs while improving schedulability for medium-priority traffic and mitigating blocking delays for high-priority traffic. The feasibility and benefits of our approach are validated on a realistic automotive TSN use case with up to 70% of reduction in TSN devices requirements compared to a full deployment.

A Cost-efficient Credit-Based Shaper Deployment Framework for Time-Sensitive Networks

TL;DR

The paper tackles the cost barrier of full CBS deployment in Time-Sensitive Networking by proposing a partial CBS deployment framework that leverages Network Calculus for schedulability verification and a cost-aware heuristic to place and configure IdleSlopes on a minimal set of legacy devices. It formalizes the CBS deployment problem as an optimization (with ILP and linearized versions) and provides a scalable heuristic to cluster CBS while preserving end-to-end deadlines, validated on automotive TSN use cases. Key contributions include the NC-based schedulability framework, the local-deadline and minimum IdleSlope formulations, and a complexity-aware heuristic with demonstrated hardware cost reductions (up to 70% fewer TSN devices and 91% fewer CBS instances) and notable delay improvements for both shaped and non-shaped traffic. The results indicate practical impact for industrial networks by enabling deterministic QoS with substantially lower capital expenditure and easier integration into existing Ethernet infrastructures. Future work will extend evaluation to larger networks, analyze load variation and interference, and explore integration with framing techniques like Frame Replication and Elimination.

Abstract

Time-sensitive networks are designed to meet stringent Quality of Service (QoS) requirements for mixed-criticality traffic with diverse performance demands. Ensuring deterministic guarantees for such traffic while reducing deployment costs remains a significant challenge. This paper proposes a cost-efficient partial deployment strategy for Time Sensitive Networking (TSN) devices within legacy Ethernet network. At the core of our approach is the Credit-Based Shaper (CBS), a key TSN scheduling mechanism. Unlike cost-prohibitive full CBS deployment, our approach selectively integrates CBS where it is most needed to enhance performance while reducing costs. Combining Network Calculus for schedulability verification and a heuristic optimization method for CBS configuration and placement, our proposal minimizes deployment costs while improving schedulability for medium-priority traffic and mitigating blocking delays for high-priority traffic. The feasibility and benefits of our approach are validated on a realistic automotive TSN use case with up to 70% of reduction in TSN devices requirements compared to a full deployment.

Paper Structure

This paper contains 17 sections, 2 theorems, 30 equations, 5 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. System Model
  4. Framework Overview
  5. Schedulability Verification Method
  6. Network Calculus Framework
  7. Sufficient Schedulability Condition
  8. Computing End-to-end Delay bounds
  9. Partial CBS Deployment as an Optimization Problem
  10. Problem Formulation
  11. Problem Linearization & Constraint tightening
  12. Cost-aware Heuristic Approach for CBS deployment
  13. Heuristic Algorithm
  14. Illustrative example
  15. Complexity Analysis
  16. ...and 2 more sections

Key Result

Corollary 1

(Left-over service curve - NP-SP Multiplexing)bouillard_deterministic_2018 Consider a system with the strict service curve $\beta$ and $m$ flows crossing it, $f_1$,$f_2$,..,$f_m$. The maximum packet length of $f_i$ is $L^{i,max}$ and $f_i$ is $\alpha_i$-constrained. The flows are scheduled by the NP

Figures (5)

  • Figure 1: Example of Time-sensitive Network Architecture
  • Figure 2: Cost-efficient CBS deployment framework
  • Figure 3: Adapted Industrial Automotive Topology from cuenot_2021
  • Figure 4: CDF for delay bounds of shaped flows, p=0
  • Figure 5: CDF for delay bounds of non-shaped flows, p=1

Theorems & Definitions (4)

  • Corollary 1
  • proof
  • Theorem 1
  • proof