Nokia Explains 6G KVIs in New Whitepaper

Back in April, Nokia published a new whitepaper on 'Technology innovations for 6G system architecture'. The paper emphasised on the six key technologies, that we have looked on this blog in the past, that will enable the essence of 6G including new spectrum bands and technologies, artificial intelligence (AI) and machine learning (ML) techniques, network-as-a-sensor, extreme networking with ultra-low latency, cognitive automated and specialized architecture. In addition, the paper also looked at the 6G vision, use cases, architecture evolution as well as new concepts to assure security and privacy.

The paper also emphasises the need for key value indicators (KVIs) in addition to key performance indicators (KPIs). Quoting from the paper:

Moving towards 6G, it is necessary to expand the fundamental network design paradigm from performance-oriented design to both performance- and value-oriented design to fully embrace the 6G vision. This requires a new class of evaluation criteria such as key value indicators (KVIs) that measure sustainability and trustworthiness, which must be understood, developed and adopted in the system architecture design evolution to 6G.

Flexibility: The 6G architecture is expected to be more flexible in various dimensions. First, the architecture must work for large-scale wide area network deployments as well as for extremely local on-premises and personal area networks. Second, in terms of function placement, the 6G architecture is expected to achieve a wide variety of latency targets and meet other requirements in a dynamic manner. For example, to support XR services, it will need to achieve ultra-low latency for specific users in constrained geographical areas and time windows. Ability to dynamically scale, for instance, to meet changing network loads, is mandatory.

Specialization: Given the variety of use cases and deployments, the 6G architecture also needs to be able to provide tailored features and functions. For example, a specific 6G sub-network, a lightweight sensor network, or some extreme networking use cases may require a specific set of functions, while other features can be omitted as they are not needed in this type of network.

Robustness and security: Users in vertical industries expect that the network service is provided in a robust, resilient and truly ubiquitous manner allowing for multi-connectivity where applicable and required. Also, we must leverage the best from different worlds, for example, facilitating the co-existence of a macro network architecture with full mobility and roaming support while seamlessly interworking with an architecture for a short-range, cell-less THz network and a non-terrestrial network architecture (NTN). A further core goal for security is to meet the strong expectations around trust and privacy.

Cloud platform: The ongoing shift to hosting of network functions in cloud platforms will continue and become broader. Deployments will, to a large extent, move from specialized telco cloud platforms to generic public, private, or hybrid clouds that can be located on-premises, in the (far) edge or in a central site. 6G will provide a uniform orchestration interface for service management of distributed clouds, complemented with segment specific abstractions, e.g., for the 6G RAN. This allows for use of specialized capabilities in the cloud nodes, such as, hardware acceleration.

Programmability: Implementations will also achieve a new level of programmability, like hardware independent programming languages such as P4, and will need to run on any cloud platform. Serverless services and functions from multiple vendors will become even more easy to integrate through open and service-based interfaces and a more modular system design. This will allow deployment of the network or instantiation of network slices with those, and only those, functions and services that are required.

Simplification and sustainability: As we expect more and more from our networks, at the same time we also must investigate opportunities for simplification of the architecture. The complexity and number of functions in the system has grown over the last generations. While more powerful zero-touch automation capabilities offer one means to cope with system complexity, the introduction of 6G also offers the opportunity to re-visit the architecture and clear out, re-design or merge functions, introduce a service based approach also for the control plane between the Radio Access Network (RAN) and the core network jointly with a distributed Non-Access-Stratum (NAS) interface, simplify signalling procedures, etc. As another example, simplification can be achieved by a tailored protocol stack, allowing for an efficient integration of customized 6G subnetworks. Simplification and providing more flexibility and dynamicity in function placement will also help to achieve sustainability goals by reducing the amount of signalling and energy consumption. A clear commitment to “design-to-sustainability” will be important for 6G.

Artificial intelligence and machine learning technology is expected to be an integral part of the 6G architecture. It is a technology that is needed to achieve the vision of a truly cognitive network that adapts itself to a variety of scenarios and deployments. AI/ML is expected to play a key role in powering the automation and optimization of the network, as well as in increasing the security of the system. The extensive use of AI/ML requires the architecture to incorporate functions and interfaces for large-scale data collection, processing and distribution, training for model refinement, and updating inference models within functions.

Besides the listed architecture design goals, there are additional aspects to be considered on our path to a 6G architecture, such as migration from 5G, deployment options to be supported, and identifying an appropriate standardization approach for 6G. The architecture needs to be future-proof, agile, cater to new and unknown applications, and consider various technology trends expected in the 6G era. It is also essential that the 6G system be able to interwork with, at least 5G, and ideally, previous generations.

One of the other important topic that the paper looked at is RAN – core redesign. Quoting from the paper again:

On the functional level, one of the major research areas under consideration are new functional splits within the RAN and between the RAN and the core network. From the RAN point of view, it can be advantageous to deploy some of the functionalities in a more centralized manner. For example, it may be more efficient to centrally perform certain tasks for a group of cells instead of doing the computation at each base station (more simple base station design) or being able to optimize network parameters for larger geographical areas (while base stations will be deployed more densely).

This trend is already visible in 5G. For instance, the central unit (CU) hosting the latency-tolerant higher layers of the radio protocol stack can cover a large service area and can be deployed in the edge or regional cloud. It works in combination with distributed units (DUs) that deal with the latency-sensitive lower layer aspects and are placed in close vicinity to the cell sites and the radio frequency (RF) unit. In this way, one CU can serve a large set of DUs.

In 6G, the modularity of the RAN and leveraging of centralized processing (e.g., with centralized function deployment being dynamically adjusted based on service requirements and network status) is not only expected to continue, but will be extended to meet the needs related to massive data collection and processing that will be an integral part of next-generation communications. Centralized functions will help realize operation, monitoring and control within the RAN as well as coordinating among RAN nodes in an ever-more optimized manner.

In addition to moving RAN functions closer to the network centre, certain RAN and core network functions will need to be pushed towards the network edges and deployed in a distributed manner. The hierarchical processing that exists in current 5G networks will not support the demanding new use cases that are defined (and yet to be defined) for 6G. Upon analysis of these use cases and their requirements, such as service latency and real-time constraints, certain RAN and core features may need to be deployed at full or almost full capacity in a localized manner, closer to the UEs. This means that the current RAN logical split and RAN and core functionalities need to be redefined so that they can be dynamically deployed in the same physical sites and share supply models and infrastructure mechanisms to support the defined use cases.

As of today, the RAN and the core network control plane are connected via the N2 interface between the CU-control plane (CU-CP) and access and mobility function (AMF) as depicted in Figure 4(a). Due to the RAN and core network functions being co-located, in this example, functions beyond the CU-CP and AMF are likely to end up being deployed in the same cloud site, while signalling remains bound to the RAN-core interface. 

To address this trend of RAN and core components being co-located, several potential solutions are currently being investigated. For future 6G systems, service-based architecture principles are expected to be applied to the RAN-core interface by defining a service-based interface between RAN and core as in Figure 4(b). This will enable a direct interface between the RAN and multiple NFs within the new 6G core.

Adapting the RAN-core interface to be service-based is not sufficient to exploit the full potential of service-based architecture principles if the non-access stratum (NAS) protocol terminates in a single point in the network. This would mean that all the signalling messages sent and received by the UE would still be using point-to-point communications, thus not fully exploiting the service-based interface between RAN and core. To leverage the full potential of SBI, distributed NAS termination needs to be considered. It would enable signalling between UE and appropriate network functions without having to cross a single point of termination every time. This design is expected to come with a range of benefits like easy and flexible adding of new functionalities to support new services, reduced signalling loads at the single point of termination, reduced signalling latency due to direct communication, and strengthened security for each termination by keeping them independent (e.g., avoiding a single point of failure) so as to enable NAS security between UE and each NF.

For a further step towards a fully service-based RAN-core signalling, a complete opening among RAN and core functions could be considered, as shown in Figure 4(c), where all RAN and core functions communicate directly with each other. Functionally, RAN and core services would still need to perform standardized and predictable operations needed for their separate domains, but core and RAN functions can provide a larger range of services that can be exposed to the other domain directly via SBIs. This needs to be evaluated -on a case-by-case basis as today’s clear separation between the RAN and the core functions has a valid justification in its separation of concerns. Applying the SBA principles inside the RAN comes with its own challenges as many of the physical network designs found in the RAN are not well suited for a service based design. Some issues, like latency, would need to be considered.

The paper can be downloaded from here.

Related Posts