Latency-Critical Networking

The network has always been driven by data, especially under the current background with rapid information technology evolutions. Latency is one of the most critical metrics in the data transmission network since it significantly affects network service quality. The demands for latency also has evolved. Not only a shorter time is required, but also a worst-case transmission guarantee should be offered. Ethernet is used in the modern network as a universal solution due to its low cost and high bandwidth. It has become more widespread in the areas such as automotive In-Vehicle Network (IVN) and industrial control network. However, following a besteffort mechanism, conventional Ethernet does not provide a latency guarantee on data transmission. The IEEE 802.1 working group has proposed the Time-Sensitive Networking (TSN) as an extension to the IEEE 802.1Q Bridges and Bridged Networks. TSN particularly addresses the deterministic low-latency requirement and high availability. Traffic scheduling paradigms are applied in the TSN network for switching and queuing operations. TSN has been widely adopted in converged networks that combine different traffic types with real-time application data. Dedicated TSN profiles have also been proposed to integrate TSN in various use cases.
This thesis investigates achieving latency-critical networking with TSN from three aspects.
1) The thesis summarizes all TSN asynchronous scheduling algorithms, compare latency, frame loss, and queue occupancy performance in a simulation environment.
2) The thesis explores implementing TSN approaches in a vehicle network, create different combinations of synchronous and asynchronous scheduling methods in the automotive gateway, studies the E2E latency of automotive application data.
3) The thesis proposes a Field-Programmable Gate Array (FPGA) design of the Asynchronous Traffic Shaping (ATS) for a switch architecture, gives an overview of the hardware scale, explores a series of hardware configurations and the impact on hardware resource consumption and data capacity. Overall, the following conclusions can be drawn from the thesis.
1) A traffic scheduler can leverage the algorithms based on the leaky bucket constraint to achieve asynchronous traffic scheduling with deterministic transmission latency.
2) In a vehicle network, ATS can achieve a comparable latency performance as the synchronous scheduling method for critical traffics with less implementation effort. 3) An ATS FPGA component can perform similar data capacity as a basic strict-priority function block while needing higher resource consumption The proposed results in the thesis show the deterministic low-latency feature offered by TSN technologies and hardware implementation concerns of the functional modules. Finally, TSN-based transmission is regarded as an effective solution to build a latency-critical network.