Saihu: A Common Interface of Worst-Case Delay Analysis Tools for Time-Sensitive Networks

TL;DR

The paper addresses the challenge of comparing worst-case delay bounds across multiple time-sensitive networking tools by introducing Saihu, a unified Python interface that orchestrates xTFA, DiscoDNC, and Panco from a single network description. Saihu automatically translates a network into tool-specific inputs, executes analyses in parallel, and exports both human-friendly and machine-friendly reports, facilitating easy identification of the smallest delay bound among methods. Through case studies on interleave tandem, ring, mesh, and Airbus-scale networks, the work demonstrates Saihu’s ability to streamline benchmarking while revealing topology- and method-dependent tradeoffs. Overall, Saihu reduces integration overhead, broadens accessibility to TSN analysis, and provides a foundation for incorporating additional tools in a consistent framework.

Abstract

Time-sensitive networks, as in the context of IEEE-TSN and IETF-Detnet, require bounds on worst-case delays. Various network analysis tools compute such bounds; these tools are based on different methods and provide delay bounds that are all valid but may differ; furthermore, it is generally not known which tool will provide the best bound. To obtain the best possible bound, users need to implement multiple pieces of code with a different syntax for every tool, which is impractical and error-prone. To address this issue, we present Saihu, a Python interface that integrates the three most frequently used worst-case network analysis tools: xTFA, DiscoDNC, and Panco. They altogether implement six analysis methods. Saihu provides a general interface that enables defining a network in a single file and executing all tools simultaneously without any modification. Saihu further exports analysis results as formatted reports automatically and allows quick generation of certain types of networks. With its simplified steps of execution, Saihu reduces the burden on users and makes it accessible for anyone working with time-sensitive networks.

Figures (12)

• Figure 1: Data flow of Saihu
• Figure 2: Device model
• Figure 3: Physical and output port network examples
• Figure 4: Human-friendly Markdown report
• Figure 5: Interleave tandem network