[Completed] [Deterministic Networking] Stability of time-sensitive networks: strategies for partial regulation

Contact: Ludovic Thomas


Time-sensitive networks are used for safety-critical systems in planes, satellites or even power plants. Their significance has increased in the last decade (autonomous cars, industry 4.0, …). While traditional public networks aim at reducing the mean service performances (ping, throughput), time-sensitive networks provide guarantees for the worst case (e.g. guarantee of a maximal latency, guarantee of a minimal throughput).

Stability is a major challenge for time-sensitive networks. Complex networks might indeed become unstable: the latency of packets can no more be bounded. This can occur when the network has cyclic dependencies (flows are interfering with each other in a circular pattern). Circular patterns are often used for redundancy in safety-critical systems. To remove cyclic dependencies in that context and thus ensure the global stability, one can implement “regulators” within all routers of the network. Regulators are hardware components that reshape the traffic [1]. They break cyclic dependencies and ensure the stability of networks, but they come with a hardware cost.

“Partial regulation” is a new strategy in which only a subset of routers acts as regulators. This strategy can reduce the cost of regulation while keeping good network performances. Questions such as “how much regulators have to be deployed?” or “where to deploy the regulators?” must be addressed.

In this project:

We will study the strategy focusing on network performances. We will develop an algorithm for identifying the regulator positions that provide the best network performances. We will focus on a cyclic network, and we will use Network Calculus [2] to compute delay bounds.

Project Goals:

  • Build an algorithm that proposes the best position for a regulator in a network with one cyclic dependency.
  • Infer heuristics for identifying the best position without requiring analysis of delay bounds.
  • Extend the previous work to a network with two cycles in order to extend the heuristics to general topologies.

Student profile

  • Python or Matlab.
  • Interest in networks, communications or real-time systems would be a plus.


  1. Latency and Backlog Bounds in Time-Sensitive Networking with Credit Based Shapers and Asynchronous Traffic Shaping Mohammadpour, Stai, Mohiuddin and Le Boudec. 2018 30th International Teletraffic Congress (ITC 30)
  2. Network Calculus Wikipedia Page