As we have seen on previous posts on the blog, for example – https://www.vamsitalkstech.com/cloud/the-three-key-use-case-areas-for-5g/– , 5G represents a revolutionary set of technical capabilities as well as enables a range of industry vertical use cases. The heart of what makes this possible is a new RAN (Radio Access Network) architecture that is not only distributed, modular but also supports open interfaces.
In this post, let us identify the core architectural components of a mobile network. While some of the terminologies can seem confusing, I will try to distill it down to the bare essentials needed to architect cloud-native applications.
The heart of 5G is the cellular network. The network supports connectivity for user equipment such as smartphones, tablets, and digital assistants. Increasingly, these will include drones, autonomous vehicles, robots, machinery and all kinds of home appliances, etc.
– The separation between the User Plane (UP) functions from the Control Plane (CP) functions. This allows both planes to evolve, scale separately thus allowing decentralized physical deployments e.g. centralized data center, cell site, edge location, or distributed location such as a remote manufacturing facility/retail store, etc. 5G networks will usher in revolutionary improvements using the above capabilities in performance, speed across industries.
– Efficient slicing by modularizing function design, e.g. to enable flexible and efficient network design. This will enable a wider range of services across distances at lower data rates.
– Enable service-based design allowing procedures that are interactions between network functions enabling their reuse. NFs (Network Functions) can interact with other NFs without using an intermediary to connect them. In fact, as we will see in later blogs, 5G platforms are designed to be cloud-native from the get-go. In use cases where distortion of waves happens in the high-frequency spectrum from physical obstacles such as buildings, weather, etc – it is advantageous to achieve a distributed architecture.
The 5G architecture includes a core network that integrates the IP-based internet with different access networks. A key design tenet is to decouple the Access Network (AN) and the Core Network (CN) by eliminating dependencies. The spec thus defines a common AN – CN interface that integrates different Access networks – both existing and new such as Wireless LANs, fixed broadband, satellite, etc. The RAN manages the radio network and ensures that QoS requirements are met for users.
Whereas predecessor RAN architectures such as in 2G,3G, and 4G were based on monolithic designs, in 5G design, the gBN (the NR node) is split between Central and Distributed Units (CU and DU). This allows flexible deployments of both hardware and software to achieve high levels of scalability. The Radio Unit is deployed closest to the towers at the Edge. The DU is deployed in numbers based on requirements for latency, bandwidth, and the needed signal strength. The DU provides real-time L1 and L2 functionality. It handles the RLC, MAC, and also sections of the PHY layer. The CU handles the RRC and PDCP layers. The gNB logical node at a minimum consists of a CU and a DU connected to the CU via Fs-C and Fs-U interfaces for CP and UP respectively. The CU handles non-real-time L2 and L3 functionality. As mentioned, a single CU manages multiple DUs over a mid-haul interface.
(Image Credit – Pipeline Magazine)
As 3GPP states, the NG-RAN architecture really builds on the foundation of 4G LTE and supports the following diverse requirements for 5G.[1]:
- Support for enhanced mobile broadband services;
- Support for Ultra-Reliable, Low-Latency Communications (URLLC);
- A single architecture to accommodate centralized, distributed, and monolithic deployments – a cornerstone of 5G, supporting the deployment of certain functions in the cloud where it is beneficial;
- In conjunction with the previous point, the possibility to fully separate Control Plane (CP) from User Plane (UP) of a centralized unit, for maximum deployment flexibility – another cornerstone of 5G, given that CP and UP scale differently. This enables the Radio Access Network (RAN) to follow the evolution (and “cloudification”) that has happened in the core network in recent years;
- Cooperation and resource sharing with existing LTE networks, to continue leveraging the operators’ installed base – a novel kind of requirement for a new Radio Access Technology;
- Flexibility to accommodate different migration strategies and “paths” from 4G to 5G, given the uncertain way in which the mobile telecommunications market is evolving;
- Time pressure to release “world class specifications” while at the same time accelerating the release process as much as possible; this effectively resulted in three release “drops” (“early drop,” “regular drop,” and “late drop”) for the same 3GPP Release.
The goal for 5G architecture is to enable vendors to develop highly flexible, integrated or decoupled hardware and software in a way that they can be mixed and matched as needed. The aforementioned distributed CU/DU model allows for better load management, higher performance and support for various QoS features for use cases such as gaming, robotics, AR/VR etc. CSPs can mix and match deployment footprints based on scenarios such as latency needs, rural/metro deployments that may or may not have access to fiber. The 5G core handles the Session Management Function (SMF) and the Access and Mobility Management Function (AMF) among a whole host of other functionality.
Now that we have covered a very high-level overview of 5G NR architecture and its distributed nature, the next post will discuss why containers and K8s are well suited to be the platform for developing & hosting these.