Top 3 challenges in cloudifying Radio Access Networks
There is an increased interest in cloud-native technologies for 5G Radio Access Network (RAN). The promise is immense, and Ericsson envisions that everything that can benefit from running in the cloud will be running in the cloud. When applying the cloud-native paradigm to a RAN it comes with challenges for the whole industry. While some of those are analogous to challenges already experienced in IT and telco network core, the RAN carries unique characteristics and requirements that increase complexity.
In the blog Pre-integration – lowering the barrier for deployment of Cloud RAN, we described some of the challenges communication service providers (CSPs) face in their RAN cloudification journey and argued for how pre-integrated solutions would lower the barrier to deploying Cloud RAN.
This blog goes further into specific and technically demanding characteristics applicable when cloudifying the RAN, such as handling disaggregation with high interdependency in RAN, secure high-performance computing under constrained resources and how to manage deployments geographically distributed from core and Centralized Unit (CU). Ericsson is working closely with its partners like Red Hat to address those, allowing CSPs to take the advantage from cloudification.
Drivers behind cloudification
With Cloud RAN, Ericsson provides a cloud-native software solution that offers increased flexibility, faster delivery of services, greater scalability in networks, and better ability to tailor solutions. Further, introducing cloud-native principles and technologies allows CSPs to employ and benefit from a common approach - from IT to their telco core network and RAN domains.
Together with its partners and CSPs, Ericsson is addressing challenges with solutions that provide value and simplify the journey for the CSPs while bringing the flexibility enabled by cloud-native and disaggregated RANs. Building on a multi-year collaboration across 20+ technology projects spanning Core Networks, Network Management, and Business & Operations Support Systems, Ericsson and Red Hat are providing concrete Cloud RAN solutions to help give CSPs greater confidence in accelerating network transformation and reducing risk.
While this blog post focuses on technical challenges, a baseline priority we should mention is realizing the potential to reduce the total cost of ownership (TCO) by cloudifying RAN, which has often previously been cited as a top desire, if not a key driver. There is an opportunity to achieve cost reduction with the use of automation, leveraging cloud economics and supply chain flexibility, as well as benefiting from the scale of general-purpose chipsets.
However, CSPs also see risks counterweighing those opportunities, including the integration costs and complexity, CAPEX (e.g., servers), and OPEX (e.g., power) related to deployment and operations of distributed RAN nodes across thousands of sites. Therefore, simplifying management of this distributed infrastructure needs automation and orchestration, which in turn requires the CSP to build new capabilities and adapt new tools. In short, we believe a fundamental transformation of how networks are built and operated will realize the TCO potential involving both CSPs and equipment/solution providers.
Cloudification of RAN: top 3 challenges
Virtualization in the telecom world is far from new. Virtualized Network Functions (VNF) comprise more than 50 percent of Mobile Core Network revenues, up from 0 percent in 2015 (Dell’Oro Group MCN Report).
Server virtualization enables running multiple operating systems on the same server using a hypervisor between virtual machines and the physical hardware. Cloud RAN is based on bare metal, and uses a containerized architecture, which allows multiple applications to run on the same operating system and same server, using a host kernel. Containers are considered more efficient, scalable, and lightweight in terms of resource overhead compared to virtual machines, which include a hypervisor layer. When building a containerized Cloud RAN solution, we face the same challenges as anyone doing cloudification. Further, realizing any modern network fundamentally comes with a broad range of difficulties the industry has addressed over time with purpose-built solutions, and those are renewed or exacerbated in the cloudification journey. Among those, we list three challenges that are particularly difficult for a disaggregated and cloudified RAN deployment, and we follow up each with concrete examples of how Ericsson and Red Hat are collaborating on engineering and validation to address them within joint pre-integrated solutions.
Challenge #1: Disaggregation with high interdependency in RAN
Cloudification and disaggregation inherently require considerably more system integration, both between vendors within a complete solution and with other network elements in a CSP’s network. Instead of partnering with one vendor who can provide the end-to-end system as in a traditional, purpose-built RAN, a cloudified RAN system could consist of multiple subsystems and vendors, resulting in a more fragmented and complex system. A CSP needs support from suppliers and partners to mitigate this complexity.
Disaggregation discards the unified node concept and introduces new challenges in life-cycle management (LCM) since functions, platforms, and hardware versions have their own release cadence. In other words, while the solution may work at deployment, the interoperability between the parts may fail every time one vendor upgrades its software or hardware. This brings another key characteristic to the CSP – the effort to ensure interoperability does not stop at “Day Zero” but extends throughout the network’s life.
Furthermore, unlike in an IT solution, there is a strong interdependency in RAN between the application and the CPU, acceleration, and networking capabilities of the underlying hardware. Access and control of these are required to come directly from the application due to the real-time processing requirements in RAN. Consequently, since these are not fully managed by the software platform, this creates more interfaces and interdependency across the stack of applications, software platform, and hardware, which in turn requires more collaboration between the vendors of the components.
In addition, disaggregation and adoption of general-purpose compute hardware bring the challenge of time synchronization to the fore. The RAN Distributed Unit (DU) requires support for synchronization protocols that work in various configurations and a higher level of clock stability. In purpose-built solutions, this was solved in hardware cascaded through the network solution. In Cloud RAN, synchronization is sourced outside the application and needs to be provided through interfaces across a multi-vendor solution.
Disaggregation also has implications on security. RAN deployments, including Cloud RAN and open RAN, in public and hybrid clouds have an expanded threat surface due to introduction of third parties in a multi-tenant environment. Each layer of the cloud stack must be protected from external and internal threat actors, who can exploit common vulnerabilities to compromise confidentiality, integrity and availability of data, and networking functions. The Cloud RAN security challenges and how Ericsson addresses them are expanded on in the blog post - Evolving to a strong Cloud RAN security posture.
Further, responsibility for security in a disaggregated model is necessarily shared among the component providers. Operators and customers face the challenge of having clear ownership of who is responsible for each layer of the stack.
Challenge #2: High-performance computing under constrained resources
The need for real-time processing and demands on the hardware keep increasing. A mid-band 5G deployment requires compute power magnitudes higher than for 4G. This processing requirement is particularly true for the RAN DU. Compared to most cloud deployments, a RAN consists of a large share of geographically distributed sites since the DU needs proximity to the Radio Units (RUs). While these “edge sites” span a large range of requirements, physical footprints, and types of locations, most edge sites are constrained in space, power, and cooling.
Additionally, the server cost and footprint need to be small in order to provide coverage and capacity while minimizing costs. Since there is a direct correlation between the system capacity and the number of cores available to the RAN application, minimizing the footprint required by the CaaS (container-as-a-service) and platform-as-a-service (PaaS) layers is also critical. Without overstating, running a DU effectively with sufficient computing under those constraints is very demanding.
From an application perspective, securing a high-performance RAN requires reliable access to underlying compute resources. For this, technologies like Data Plane Development Kit (DPDK) and Single Root-input/output virtualization (SR-IOV) are central. While those are not unique to cloudified RAN, their application in a distributed and highly time-sensitive environment creates incremental requirements for the CaaS layer.
Challenge #3: Managing deployments geographically distributed from core and Centralized Unit (CU)
Though best practices are established for managing cloud deployments in centralized environments, there is still a journey ahead to fully adapt and improve those practices when deploying in the far RAN edge. Edge sites add to the complexity of cloudifying RAN since they often have limited spare bandwidth. Compared to purpose-built basebands that come with fully pre-integrated software and centralized core deployment with considerable bandwidth, this type of Cloud RAN site may prohibit rapid download of software for deployment. This creates a tough requirement for Day Zero deployment and configuration of the node, as well as for efficient scratch install if the node becomes unresponsive. Looking at post-Day Zero, single server deployment adds complexity. Practices around rolling upgrades require a complete Kubernetes cluster (i.e., at least three master nodes), generally not available or desirable in the network edge.
Cloudification and disaggregation also require innovation around service availability and reliability, which is the single most important characteristic of mobile networks. Purpose-built RAN has been designed and built specifically for these requirements to meet the KPIs set for mission and business-critical services. While cloudifying RAN brings the opportunity to leverage the investments and learnings from the IT domain, it also brings the challenge of achieving telco-required network resilience from hardware and software designed for very different environments. Particularly in the far edge, ensuring uptime cannot rely mainly on the redundancy of nodes in the site or across deployments.
Ericsson is partnering in Cloud RAN to bring solutions to the challenges
Ericsson is innovating and collaborating across the ecosystem to find the best possible answers (and beyond). While not an exhaustive list, let us provide a view of some concrete paths we run down to bring solutions to CSPs, sharing examples of how Ericsson is working with Red Hat to address the challenges.
Addressing challenge #1: Disaggregation with high interdependency in RAN
Ericsson offers Cloud RAN on Ericsson Cloud Native Infrastructure Solution (CNIS) as a complete pre-integrated solution. To offer customer valuable flexibility and choice, Ericsson also brings offers for multi-vendor deployment and software modularity. To ensure that the multivendor solutions meet expectations on performance, time-to-market, scale, and quality, Ericsson and its partners pre-integrate at the product level and jointly perform extensive testing and validation of those solutions across the hardware and software stack. Beyond the deployed solution’s testing and validation, life-cycle management within the stack is also orchestrated between Ericsson and the partners.
Ericsson has collaborated with Red Hat since 2019 to evolve, validate, and test the interoperability between Ericsson Cloud RAN products and Red Hat OpenShift . The Cloud RAN effort rests on a common and proven methodology to secure the solution’s quality and performance, as well as R&D efficiency for faster time-to-market. The joint development and iterative testing are greatly supported by Ericsson’s test harness. This allows Red Hat to continuously test how its hybrid cloud platform performs together with Ericsson’s network functions without the need to set up the full system. The test harness introduces a series of test cases, including onboarding, security-focused practices, resilience, and automation, to validate the consistent performance of RAN-specific functionality. Also, it validates the stability of the Containers as a Services (CaaS) platform and underlying hardware infrastructure drivers and components like CPU, accelerator, and network interface cards.
Further, Ericsson and Red Hat are collaborating to better support LCM for CSPs by offering interoperability between Ericsson functions and Red Hat OpenShift beyond Day Zero throughout the solution’s life cycle. Building on jointly defined reference configurations and shared processes around release management and testing, Red Hat and Ericsson work to ensure the integrity of each release of the solutions while supporting the release cadence of the comprised components to indicate interoperability between Red Hat OpenShift and Ericsson Cloud RAN software functions.
Regarding security, Ericsson is working across the functionality of applications and solutions, as well as processes. Ericsson Cloud RAN secures Open RAN deployments by striving toward a Zero Trust Network Architecture (ZTNA) to mitigate external and internal threats. Significant investments across the development organization have been made to achieve strong network security. Ericsson is building a Cloud RAN security solution and security assurance processes based on experiences from purpose-built RAN deployments at scale and collaborates with leading cloud platform and hardware vendors. Red Hat and Ericsson collaborate to align the level of security functions in our respective products to enhance reliability and safety measures.
The Cloud RAN application provides built-in functionality to protect data on its internal and external interfaces. On a solution level, Ericsson brings established trust through mutual authentication, automation and secure configurations, and data streaming to enable threat detection and automated configuration compliance.
To address the shared responsibility challenge, Ericsson employs the Cloud Shared Responsibility Model as a helpful tool for stakeholders to determine who has security responsibility at each layer of the cloud stack. In addition, Ericsson’s Cloud RAN offering passed the independent NESAS audit, making it fully compliant with the security requirements defined by global standards organizations 3GPP and GSMA.
Addressing challenge #2: High-performance computing under constrained resources
For a DU deployment on the edge close to the antenna site, typically with constrained resources, we see CSPs typically deploying a single node to ensure maximum resource utilization to strike a balance between performance and costs. Consequently, the Cloud RAN solution must optimize the platform’s use of CPU core cycles to leave processor cycles available to the application.
To support this typical use case, Ericsson and Red Hat have developed full RAN DU deployment within a single node. We jointly push and iterate to minimize and best allocate the resource needed between the Ericsson Cloud RAN network functions and Red Hat cloud platform to get the most out of the hardware configurations, including RAN Layer 1 acceleration. Also, for this purpose, we continuously jointly update the Test Harness and solution roadmaps to make sure the core, memory, and power utilization will not deteriorate but rather improve over time.
Addressing challenge #3: Managing deployments geographically distributed from core and Centralized Unit (CU)
There is an inherent problem with upgrading and ensuring resilience for single server deployments. To address this, Ericsson and its partners collaborate with CSPs on network planning to explore different methodologies, including planning for cross-coverage to support both upgrades and resilience. Ericsson and Red Hat collaborate to synchronize release and upgrade cycles to improve predictability and minimize upgrade windows. Further, Ericsson provides strong built-in resilience of the application software which supports CSPs’ overall network resilience.
Ericsson’s partnerships expand beyond the specific reference configurations to also address the operational challenges of geographically distributed deployments, focusing, for example, on bandwidth constraints. This may involve different types of servers’ pre-staging to be deployed as well as collaboratively working with Red Hat to minimize the image size for upgrades and deployments and, in turn, their download times to meet the CSPs’ expectations. Beyond that, we collaborate on the processes’ design and automation to efficiently light up or upgrade the solution on the site.
By working with leading CSPs on their specific networks, Ericsson and Red Hat collaborate on both solutions and operations for managing geographically distributed deployments. We believe this is one of the major opportunities to innovate in Cloud RAN.
The value of cooperation
As we said at the beginning, we envision that everything that can benefit from running in the cloud will be running in the cloud. To realize that vision with the best-in-class performance Cloud RAN, the value of a collaborative ecosystem becomes imperative. We are still only at the start of our journey. With partners like Red Hat, Ericsson is paving the way toward the vision that CSPs can deploy cloud-native networks virtually everywhere, on any cloud and server platform.
Like what you’re reading? Please sign up for email updates on your favorite topics.Subscribe now
At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.