6GSMART-EZ: A centralized control plane for Time-Sensitive Networking (TSN)

Time-Sensitive Networking (TSN) aims to group a set of IEEE 802.1 standards to provide strict Quality of Service over Ethernet networks. These standards revolve around four main quality-of-service aspects:

• Time synchronization: TSN networks provide time synchronization using 802.1AS (gPTP, generalized Precision Time Protocol).
• Bounded low latency: TSN networks allow bounded latencies through standards such as IEEE 802.1Qbv (Time-Aware Shaper), which divides time into time slots and grants the transmission of each traffic class during the assigned time slots, or IEEE 802.1Qav (Credit-Based Shaper), which is similar to a leaky bucket and allows transmission when credits for each traffic class in a queue are non-negative.
• High reliability: It is achieved through IEEE 802.1CB (Frame Replication and Elimination for Reliability), which implements the sending of duplicated frames over multiple paths, or IEEE 802.1Qca (Path Control and Reservation), used to provide bandwidth reservation and configure paths for streams.
• Resource management: It is performed through IEEE 802.1Qcc, which defines the control plane architecture in TSN networks.

A centralized TSN control plane

Following the IEEE 802.1Qcc standard, the TSN control plane comprises two main modules. Firstly, the Centralized User Configuration (CUC) is responsible for discovering end stations and grouping the stream requirements. Also, the CUC is in charge of sending the collected information using the RESTCONF protocol to the Centralized Network Configuration (CNC), which discovers the network topology, calculates the optimum schedule considering both topology and stream requirements, and applies this schedule to TSN switches by means of the NETCONF protocol.

Therefore, a typical flow of the events in a Time-Sensitive Networking scenario would begin with the end devices (Talker, responsible for sending traffic, and Listener, which receives traffic) notifying the CUC with their intention of establishing a data transmission. This would be done by sending the stream list and their requirements (frame size, maximum latency allowed, periodicity, destination…). At the same time (or even before), the CNC would discover the network topology to finally receive the information from the CUC and create an appropriate schedule that would be transmitted to the TSN switches in the last step.

IEEE 802.1Qcc also considers two extra architecture models aside from this one (the fully-centralized model), which are the fully distributed model, where there do not exist any control plane modules (CUC nor CNC) and configuration is transmitted along the topology from node to node, and the centralized network / distributed user model, where there does not exist any CUC, and stream requirements are communicated to the CNC from the TSN bridges connecting the end stations.

6GSMART-EZ control plane for Time-Sensitive Networking

As part of the 6GSMART-EZ project, UPC has developed a Time-Sensitive Networking control plane that performs as both CUC and CNC and is based on a microservices architecture (as shown in the diagram below), using Docker, Python and RabbitMQ. Therefore, the diverse functions of the CUC/CNC (topology discovery, gathering stream requirements…) are divided into containers, which provide the flexibility required to deploy microservices into different machines according to its demands in terms of computing resources. In this case, we found that the ILP Calculator module in charge of calculating the scheduling was the most resource-demanding microservice and run times were optimized when deployed on machines with high computational resources. Consequently, the modularity brought by this strategy seems to be a clear advantage compared to monolithic approaches that limit scalability.

More details and results can be found in this publication: Agustí-Torra, A.; Ferré-Mancebo, M.; Orozco-Urrutia, G.D.; Rincón-Rivera, D.; Remondo, D. A Microservices-Based Control Plane for Time-Sensitive Networking. Future Internet, 2024, 16, 120.

Author: Marc Ferré Mancebo, Researcher at the Design and Evaluation of Broadband Networks and Services (BAMPLA) research group of the Universitat Politècnica de Catalunya (UPC)