RAN virtualization disaggregates software from hardware and makes the RAN software-defined and programmable. Once a network service transitions to cloud, adopting the cloud principles of automation, agile operations, and cloud scale, the transition tends to stay sticky and stay in the cloud. In other words, telecoms networks are becoming like IT networks. And like IT networks, they are also beginning to move to the cloud.
Cloud computing is a model for enabling ubiquitous, on-demand network access to a shared pool of computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal manual management or interaction. Although not a new concept, embracing of cloud principles by communications service providers (CSPs) is still a recent development, but one that is critical as a complement technology to manage the increasing complexity of networks and services.
Cloud RAN allows network to be realized in software, by helping CSPs take advantage of cloud-native architectural principles and enabling RAN functions to be deployed as microservices in containers over bare metal servers, supported by orchestration technologies such as Kubernetes. In such an architecture, apps run in isolation (containers) but could share an operating system (OS) where everything each application needs such as configuration files, libraries and dependencies are encapsulated in its container. This approach is more supportive of DevOps principles and application innovation, and enables faster time-to-market, greater flexibility, and service agility.
There are several key benefits to the RAN functionality being software-defined, and cloud-native.
- Cloud RAN allows hardware and platform infrastructure harmonization across the network allowing for greater automation.
- Automation of management and adoption of cloud-native CI/CD principles adds new levels of agility and flexibility that promises to create significant operational performance and efficiency gains
- Software-based programmability opens the RAN environment for broader innovation and allows to tailor the behavior and performance of the RAN through applications with access to multi-domain information.
- Cloud RAN functions can be deployed, scaled, and managed consistently across different data center sites- far edge, edge, and core – thus supporting multiple network architecture models and faster deployment times.
All this sounds great, but this is of course much simpler to conceptually capture than to deploy and manage in the field for the CSPs. Typically, CSPs have an existing network infrastructure which must evolve towards cloud-based RAN, and so need an achievable, pragmatic, phased-in approach.
In addition, implementing cloud-native RAN with the new automation and operational tools whilst retaining a holistic system with existing network infrastructure and manage them in a cohesive and consistent fashion is a complex undertaking. While physically integrated RAN nodes are locationally fixed once deployed in their sites, locational independence is a defining principle and cornerstone in Cloud implementations. Thus, in Cloud deployments, RAN is no longer conceived in terms of ‘nodes’ in fixed locations; instead, network resources and functionality can be locationally distributed as software functions and composed dynamically into node components. This leads to a shift in paradigm in how the node functions in cloud are managed alongside physical network functions in a consistent way, how they interact, updated, life cycle managed etc. RAN processing also has quite specific requirements including stringent latency demands compared to IT applications and core network workloads, which need to be accommodated.
So, what might the RAN pathway to the cloud look like?
Firstly, it is vital to identify the services, use cases and deployment scenarios—urban/rural, indoor/outdoor, for example—being targeted to decide the right starting point for the journey. This typically starts with decisions on the hardware, cloud platform and the RAN applications that will be hosted on it, as well as ensuring that the cloud infrastructure will be compatible with the needs of existing customers and site assets and transport is ready to support the Cloud RAN requirements. A key aspect that cannot be over-emphasized enough is the fact that Cloud also requires new ways of managing the network, and new tooling to automate, orchestrate and troubleshoot the network. Last but not the least, cloud introduces new security challenges that the operator must deliberately plan, design, and prepare for.
Once the direction is established and the solution design is complete, the ‘cloudification’ of the RAN can begin, by migrating selected 5G RAN network functionality as containerised functions to commercial off-the-shelf (COTS) servers and accelerators. 3GPP Release 15 5G network architecture allows the baseband unit, or BBU to be split into separate elements as Centralized Unit (CU) and Distributed Unit (DU). CU is further split into User Plane (CU-UP) and Control Plane (CU-CP). Starting the Cloud RAN journey with CU offers a low risk starting point for the CSPs. This also enables distributed cloud-based edge deployment of 5G RAN CU-UP and 5G core User Plane Functions (UPF), giving flexible deployment options. Cloud based deployment of Centralized Unit control plane (CU-CP) creates additional locational and deployment flexibility.
The next step in the journey is cloud-native deployment of the distributed unit (DU) functions containing latency-sensitive lower layers of baseband processing. The innovation doors are now open for RAN Programmability and App based innovation to follow. This not only allows 5G to become a powerful platform for application innovation but also paves way for single pane management of services across Cloud RAN, transport, 5G core with a harmonized cloud infrastructure underlay.
Evolving the mobile network to Cloud RAN is a multi-year journey, in which it will be deployed in live networks co-existing with and supporting multiple radio access technologies, the majority of which will continue to be operating as physical functions. As with all network transformation journeys, this too must be undertaken without impacting the quality of experience for users.
To build and integrate Cloud RAN demands robust understanding of the entire telecom stack as well as network evolution capabilities and skills, so it is critical that CSPs choose strong technology solutions partners with a balanced and pragmatic approach towards Cloud flexibility, user experience, innovation, and alignment towards 3GPP architecture.
Performance demands for 5G use cases that demand ultra-low latency, high reliability, slicing, and traffic steering means that the cloud infrastructure also needs to evolve to handle the highly demanding new RAN applications. To meet the performance needs to effectively address new use cases and rapidly on-board new apps and services, cloud networking is evolving towards edge and far edge parts of the network all the way to on-premises and in-building deployments.
The cloud journey will be a long and complex one and the pathways are being built now. The timing to start the cloud journey will differ between CSPs but the starting point is clear: It is in realizing 5G as cloud-native functions, designed, built, and managed for the cloud and building for its seamless co-existence with existing multi-generational integrated network infrastructure.
