TSN Distributed Clock : An analysis of relationships between network configuration parameters and the resulting precision of time synchronization

Abstract: In real-time systems spanning a network, there is a need for deterministic communication. The best-effort approach which most of the Internets traffic follows is not suitable for this area since it does not guarantee packet delivery within a deadline and there is also no accurate measure of when the packet was sent. The network core and edge entities such as routers and hosts do not have any concept of time in normal networking, making real-time constraints more difficult to enforce. Time Sensitive Networking is a set of standards, all of which are related to solving the problem above. The most central of these standards is IEEE 802.1AS which defines the generic Precision Time Protocol that specifies how all the nodes of a network should synchronize their clocks to one master clock, giving them a common perception of time. This standard is a prerequisite for some of the other standards in the suite, for example, the 802.1Qbv standard defining a Time Aware Sharper which provides bounded latency for time-critical traffic. A common perception of time is also by itself needed by applications that have to orchestrate actions, with temporal relations to each other, across a network. These applications can be found within areas such as industrial automation and vehicular control systems. The problem that this thesis explores is how the precision of time synchronization of a Time Sensitive Networking (TSN) solution depends on variables in the network such as configuration, topology, and external factors. To find the correlation between the parameters and the precision of the time synchronization, several experiments have been conducted. The experiments were performed on a simple network of hardware components constituting a physical test bed and an oscilloscope was used to probe the clocks if its nodes and extract measurements. Our findings indicate several relationships between the tested parameters and the synchronization precision. The biggest conclusion we can make from our study is that the IEEE 802.1AS standard does not rely on the support of other standards to achieve sub-microsecond results when there is a best-effort traffic load on the network. The manipulated configuration of the standard has given results that in general coincide with the expected behavior. Finally, the data gathered on different topologies, that were tested showed no significant trends regarding the precision.