More flexible 5G slicing: Orange and Ericsson rely on AI

Contents

French telecommunications provider Orange and network equipment supplier Ericsson want to advance 5G slicing in Europe with a jointly developed technology. It is based on AI to both improve network utilization and increase the performance of 5G end devices.

Orchestration is carried out via Orange's “Service Orchestration and Assurance Platform”, which is designed to enable dynamic end-to-end slice management, for example for private mobile networks, geo-fencing or fixed wireless access (FWA). The basis for this is their network management software “AI-Network-Brain”, for which Orange promises modular expansion and adaptation to different customer groups.

Typical application scenarios include critical infrastructures, predictive maintenance, edge AI, robot control, autonomous vehicles, drone inspection and real-time video analytics. Orange promises guaranteed upload rates and latency times of less than five milliseconds. This should practically create a fiber-optic-like mobile connection via 5G.

Orange refers to this as Quality of Delivery (QoD), which goes beyond QoS (Quality of Service). This also includes commitments regarding monitoring, latency, jitter and reliability – i.e., more than just bandwidth. In France, Orange already offers corresponding 5G contracts for business customers, for example 350 GB with VoNR and slice security.

Network slicing: a slice of the big cake

With network slicing, part of the network capacity is “carved out”. Technically, this works in such a way that all slices run on the same physical network, but are logically isolated. Each slice has its resources: radio network (RAN), transport network and 5G core. The advantage of this is that an individual set of bandwidth, latency, availability or SLAs can be configured for specific requirement profiles or user groups. This is structurally similar to virtual machines.

Deutsche Telekom presented 5G network slicing two years ago. This solution is specifically geared towards live TV production, for example for reporters and cameramen on location. It is based on the Telekom core network with Mavenir MDCA and is orchestrated via customer interfaces. This includes dedicated QoS slices and CAMARA Open APIs to enable high-quality HD transmission. The product is now on the market as a premium service for media companies (TV, streaming) with QoS level specifications.

Vodafone also supplies 5G slicing solutions with SLA technology. However, here too, the offering is aimed at a small range of applications: events, live sports events with dedicated slices for uploads and streaming as well as a few vertical applications, for example in the context of private industrial networks. O2 has had a corresponding offer since October 2023, which is mainly aimed at business customers. There are now also plans to use it for private customers in the areas of gaming, smart home or as a tracker with health and security functions.

Previous slicing too rigid?

All these slicing offers have one problem in common: you have to specify exactly what is needed beforehand – and then delete the slice again, as it remains in place even if it is no longer needed. This setting up and releasing of a slice can also be program-controlled via API, which is mainly used in an industrial environment, for example when a mobile video circuit is required for a certain period of time. But here, too, the slice remains in place until it is switched off again by the program or after specified time settings. Until now, this has been more of a theoretical issue because in the use cases mentioned above, the corresponding load is predictable and relatively constant.

But now a new application is coming onto the market: the mobile use of AI in the business environment. This not only means more data traffic, but also a huge jump in bandwidth requirements. This is where previous slicing becomes a problem because demand is unpredictable and only occurs for a short period of time. This requires a fully automatic, dynamic allocation of slices based on the current requirements.

“Intent-based” fully automatic slicing

This is precisely where the new Ericsson technology comes in. “Our intent-based solution automates the creation and replication of 5G slices,” according to the press release. This means that the system should recognize where a large demand suddenly arises and then set up a suitable slice according to the situation and other known parameters. As soon as the demand is no longer present, the slice is automatically released again. According to Ericsson, the network can thus be used more flexibly and more subscribers can receive better performance at the same time.

As an example application, Orange showed the use of AI for video generation on a cell phone. This suddenly requires a lot of bandwidth with low latency. However, as soon as this application is over and the demand is no longer there, the slice is released again. This new technology is not just limited to AI on smartphones, but could also support industrial applications such as mobile machine and robot control, autonomous vehicles or video analytics.

Ericsson and Orange can not yet provide any references. Two use cases are known: The original test scenario in Belgium, which took place as part of a drone deployment in real time. This use case was presented at the MWC congress in February on the experimental Pikeo network, the Orange Telco Cloud and AWS outposts. And then there is another application at Schneider Electric, where a solution for industrial processes and automation was created in collaboration with Nokia and Ericsson.

(mho)