Abstract
1 ABSTRACT Before the 5G era, the wireless networks used for professional uplink streaming scenarios such as media production were either the best-effort networks provided by local mobile network operators (MNO) with cellular bonding, or the use of proprietary wireless link solutions, both analogue and digital, using private mobile radio as well as licensed spectrum (for example line-of-sight point-to-point microwave links or DVB-T based COFDM solutions). With 5G, the cellular technologies offered a broader link bandwidth, providing much higher link speeds than earlier cellular technologies, opening the door for a more interoperable and scalable ecosystem.
To cope with the non-guaranteed quality of service (QoS) provided by the earlier mobile networks, a cellular bonding based ecosystem has been successfully developed, enabling a camera equipped with multiple modems to split the traffic and fine-tune the rate allocation between multiple aggregated links. However, this still treats networks as bit pipes at the expense of additional latency and a sub-optimal use of networks resources. With 5G, non-public networks (NPN) were specified to address industry-specific connectivity needs including media [10]. Now a media production team can reserve capacity, e.g. leveraging network slicing from an MNO for a nomadic event and then access a wide range of advanced features to optimize its media production workflow. It can also deploy its own dedicated network to provide local area connectivity. Non-Public Networks (NPNs) can be deployed in multiple flavors [1-3], known as Standalone-NPN (S-NPN) and Public-Network-Integrated-NPN (PNI-NPNs). While the two options allow creation of private networks and provide access to a guaranteed capacity and QoS, they differ in multiple aspects, including spectrum-licensing [5], device-onboarding [4], device management and APIs. In addition, traffic steering switching and splitting (ATSSS) can be a complementary solution to NPNs to offload traffic to locally available non-3GPP access (e.g. WiFi).
For consumer uplink streaming services running on smartphones, advanced cellular bonding technologies are not available, since majority of smartphones leverage only one Modem for communication. Thus, the Quality of Experience (QoE) for the viewer is mostly unpredictable due to the best-effort networks the users are connected to. While leveraging slicing or prioritized connectivity for such usage may not be commercially available nor really considered by consumers, other functionalities coming from the 5G toolbox can be leveraged and exploited by the application to enhance QoE. These functionalities are offered as part of the 5G Media Streaming (5GMS) toolbox developed by 3GPP SA4 which provides an architecture and a set of functions enabling more reliable video streaming sessions for uplink and downlink scenarios.
5GMS covers multiple aspects spanning from architecture [6], multimedia profiles [7, 8] to protocols [9]. It provides functionalities to support content management (e.g. ingest, hosting and publishing) but also traffic management through network assistance and dynamic policies. In addition, metrics/QoE reportin and edge capabilities are supported. While these aspects have been well studied for downlink [11], the optimization of uplink media workflows leveraging 5GMS on today's best-effort networks is tackled.
This talk focuses on features enabling uplink streaming scenarios to be enhanced when deployed over 5G. More specifically, the on-boarding of devices into slices and how their traffic can be precisely managed thanks to 5G APIs is depicted. The media synchronization mechanism, essential to any production workflow, is explained, exposing the different possible technical approaches and their limitations. The consumer-centric use-cases are covered, addressing media production and streaming based on Smartphones. A deep dive into 5GMS functionalities essential to enable high QoE uplink streaming is provided. Specifically, we explain how network assistance and metrics reporting can be leveraged to smoothly adapt the streaming session, coping with possible QoS fluctuations. Finally, edge computing and media processing offload is investigated.