Ethernet Fronthaul and Time-Sensitive Networking for 5G and Beyond Mobile Networks

Ethernet has been proposed to be used as the transport technology in the future fronthaul network. For this purpose, a model of switched Ethernet architecture is developed and presented in order to characterise the performance of an Ethernet mobile fronthaul network. The effects of traditional queuing regimes, including Strict Priority (SP) and Weighted Round Robin (WRR), on the delay and delay variation of LTE streams under the presence of background Ethernet traffic are investigated using frame inter-arrival delay statistics. The results show the effect of different background traffic rates and frame sizes on the mean and Standard Deviation (STD) of the LTE traffic frame inter-arrival delay and the importance of selecting the most suitable queuing regime based on the priority level and time sensitivity of the different traffic types. While SP can be used with traffic types that require low delay and Frame Delay variation (FDV), this queuing regime does not guarantee that the time sensitive traffic will not encounter an increase in delay and FDV as a result of contention due to the lack of pre-emptive mechanisms. Thus, the need for a queuing regime that can overcome the limitations of traditional queuing regimes is shown. To this extent, Time Sensitive Networking (TSN) for an Ethernet fronthaul network is modelled. Different modelling approaches for a Time Aware Shaper (TAS) based on the IEEE 802.1Qbv standard in Opnet/Riverbed are presented. The TAS model is assumed to be the scheduling entity in an Ethernet-based fronthaul network model, located in both the Ethernet switches and traffic sources. The TAS with/without queuing at the end stations has been presented as well. The performance of the TAS is compared to that of SP and WRR and is quantified through the FDV of the high priority traffic when this contends with lower priority traffic. The results show that with the TAS, contentioninduced FDV can be minimized or even completely removed. Furthermore, variations in the processing times of networking equipment, due to the envisaged softwarization of the next generation mobile network, which can lead to time variation in the generation instances of traffic in the Ethernet fronthaul network (both in the end-nodes and in switches/aggregators), have been considered in the TAS design. The need for a Global Scheduler (GS) and Software Defined Networking (SDN) with TAS is also discussed. An Upper Physical layer functional Split (UPS), specifically a pre-resource mapper split, for an evolved Ethernet fronthaul network is modelled. Using this model and by incorporating additional traffic sources, an investigation of the frame delay and FDV limitations in this evolved fronthaul is carried out. The results show that contention in Ethernet switch output ports causes an increase in the delay and FDV beyond proposed specifications for the UPS and other time sensitive traffic, such as legacy Common Public Radio Interface (CPRI)-type traffic. While TAS can significantly reduce or even remove FDV for UPS traffic and CPRI-type traffic, it is shown that TAS design aspects have to carefully consider the different transmission characteristics, especially the transmission pattern, of the contending traffic flows. For this reason, different traffic allocations within TAS window sections are proposed. Furthermore, it is demonstrated that increased link rates will be important in enabling longer fronthaul fibre spans (more than ten Kilometres fibre spans with ten Gigabit Ethernet links). The results also show that using multiple hops (Ethernet switches/aggregators) in the network can result in a reduction in the amount of UPS traffic that can be received within the delay and FDV specifications. As a result, careful considerations of the fibre span length and the number of hops in the fronthaul network should be made.