As the next generation of wireless technology, 5G significantly differs from previous wireless generations. While the previous wireless generations were designed to connect people to people as well as to the internet, 5G has a broader scope. It connects things to people, to the internet and to things. It also caters to a wide range of industry verticals, meeting the requirements of flexibility, latency, cell density, spectrum of radio frequencies and data rates. Consequently, networks are evolving, utilising software-defined networking (SDN), network function virtualisation (NFV) and cloud-native architectures to enable the disaggregation and virtualisation of primary functions. This leads to the separation of the control plane and the user plane, and introduces capabilities such as network slicing and mobile edge computing.

Disaggregation and virtualisation

The 5G architecture is essentially designed to take advantage of cloud-native concepts – the ability to leverage self-contained functions within and across data centres (the cloud), communicate in a micro-services environment, and work together to deliver services and applications.

Disaggregation and virtualisation are two key elements of 5G cloud-native architecture. The use of SDN and NFV basically allows the disaggregation and virtualisation of many telecommunications and mobility functions such as S/P-GW (serving/public gateway), MME (mobility management entity), RAN CU/DU (central unit/distributed unit), TDF (traffic detection function), IP routing and Ethernet encapsulation/switching. These functions are hosted as software services and dynamically instantiated in different parts of the network segments. Thus, the overall 5G network is designed to be software configurable.

Control and user plane separation is the concept of disaggregation that allows the two planes to exist on separate devices or in separate locations within the network. For example, in the core, the user plane functionality is separated from the control plane functionality in the serving gateway, public data network gateway (P-GW) and TDF functions. Separating the control plane from the user plane allows the two planes to scale independently, without having to augment the resources of one plane when additional resources are required only in the other plane. The separation allows planes to operate at a distance from each other as they are no longer required to be co-located.

From a functional disaggregation perspective, the service, control, data and management plane separation is already happening on transport systems using SDN. As per the direction provided by the latest standard specifications for long term evolution-advanced (LTE-A) and 5G, functional disaggregation also takes place on the mobile network element layer. For example, 5G RAN is disaggregated into CU and DU functions, and within the CU, they are disaggregated into CU-CP and CU-DP. When all data plane functions of different network elements are disaggregated, the data plane is distributed using a consolidated set of protocols. The data plane functions could either be realised via a virtual network function construct multi-access edge computing platform or as a programmable application-specific integrated circuit construct (programmable transport underlay). The transport control plane and data plane protocols are expected to consolidate and simplify as network systems adopt a cloud-native construct.

In RAN, cloudification allows the disaggregation of the remote radio unit (RRU) from the baseband unit (BBU). By separating these functions, it becomes possible to create a pool of BBU resources that supports several distributed RRUs. This is referred to as a C-RAN, or CloudRAN, where elements of the RAN can be centralised and implemented in the cloud as well, enabling a more efficient use of RAN resources. It also creates some challenges, such as the need for fronthaul connectivity between RRUs and BBUs. This challenge is being addressed by architectures that define the splits at different locations in the RAN,  and have trade-offs between bandwidth requirements and the ability to centralise resources.

This has also led to initiatives such as xRAN. In 5G, fronthaul will move away from CPRI-centric interfaces to a type of packetised fronthaul. Multiple standards definition organisations are releasing specifications for packetised fronthaul for usage with 5G and LTE-A Pro networks. These include eCPRI 1.0 from CPRI.info; the Institute of Electrical and Electronics Engineers’ 1914.3 RoE (radio over Ethernet); and xRAN.org’s xRAN Fronthaul 1.0.

All these specifications are focused on realising the fronthaul in the physical layer of the RAN. In particular, xRAN Fronthaul 1.0 drives the possibility of open radio equipment and radio equipment controller, where multivendor radio controllers and active antennas could interoperate; and provides an open ecosystem of radio control applications such as hybrid load balancing, eICIC and self-organising networks.

 

Excerpts from a white paper, “5G at the Edge”, released by 5G Americas in October 2019