Openness Matters: The Move Towards Containerization
So far in my Openness Matters blog series, I have outlined the various factors that are prompting service providers to adopt open standards, as well as the ways they can deliver a seamless and optimized end user experience through the convergence of wireless and wired networks. In this edition, I want to investigate the role of containerization in exponentially increasing the flexibility of service providers to deploy and manage new services.
Communications service providers (CSPs) have already started to design and deploy cloud native network functions with the introduction of commercial 5G standalone solutions. CSP legacy networks such as LTE/4G already have certain functions implemented as virtualized network functions (VNFs). These functions will need to gradually evolve towards containerized network functions (CNFs) to minimize operational management overhead and to drive towards uniformity in their network evolution. Although 5G standalone deployments are still in their infancy due to the overall ecosystem maturity, the pressure is building up for CSPs to transform their networks to stay ahead and remain competitive in the industry.
As the name implies, 5G cloud native networking functions are built to run flexibly in public, private and/or hybrid cloud environments. These functions need to be able to take full advantage of the cloud computing benefits including resiliency, scalability, observability, and adaptability. Also, the addition of an intelligent automation layer is critical to the success of the cloud native networking evolution as it simplifies the overall lifecycle management of monolithic software applications that are broken down into “microservices” as components or building blocks running on the same container.
Containerization is a way of putting all the essential elements of an application – the software code, libraries, frameworks, and various other dependencies – and bundling them all together into an independent “container.” Not only are these containerized applications fully functional and able to run regardless of the operating system or platform, but they are also highly portable. Because they are decoupled from the proprietary control and management software, containerized apps can easily be deployed in any networking environment.
The containers that shape up an application function run as independent software entities that can be instantiated, moved across cloud environments, elastically scaled in/out, and de-instantiated in a rapid manner. CSPs can benefit from cloud agility and efficiency by running the 5G application functions as containerized microservices. Migration from monolithic physical network functions (PNFs) and VNFs to CNFs is a major shift in the digital network transformation for CSPs; and these CSPs must establish a solid baseline for 5G networking solutions.
While there has been a strong push in the communications service provider community to embrace cloud native digital network transformations with the advances in next-generation information communication technologies, the fundamental question that one needs to ask is “does every network element and/or function within a telco network need to be cloud-native?”
The answer is a clear “NO” because service providers must deal with their legacy network infrastructure without any disruptions to their day-to-day operations, and these networks could be monolithic with physical networking functions providing revenue generating services today. Several global CSPs will continue to run their legacy networks (4G/3G/2G) using a combination of conventional physical and virtual network functions as they gradually evolve their 5G network expansion plans.
The software-defined disaggregation and containerization of new 5G application functions across simultaneous radio access and core networking domains as well as the underlying transport connectivity adds to the overall complexity of modern cloud native network transformation in terms of analysis, design, implementation, integration, deployment, and lifecycle management. Hence, it’s not mandatory that every single network function across the multiple networking domains has to be cloud native from day one.
Software re-factoring requires significant investment in terms of R&D costs to address simultaneous changes across multiple domains and deliver the same level of end-to-end network functionality, performance, capacity, reliability, availability, etc. While vendors tend to implement the containerization of 5G networking functions in different ways based on their domain expertise, timelines and market demands, such changes must be implemented in a coordinated manner by CSPs across their network segments working with multiple vendors that will deliver the best results for their infrastructure investments.
5G Core Network
In terms of the criticality of the mobility network infrastructure cloud native transformation, the 5G Core network is at the forefront given its maturity in terms of 3GPP standards definition of the various application functions as well as its interworking design with legacy LTE-EPC core network functions. While operators have embraced virtualization of their critical EPC core networking functions and disaggregation (via control and user plane separation) over the past decade, not all network functions may have gone through such transformation due to the legacy dependencies associated with 3G interworking, regulatory services that depend on higher levels of reliability associated with incumbent physical network functions, and the anticipated ROI in terms of implementation complexity, ease of operational use and associated timelines. The fact that the 5G Core network architecture is based on a unified service-based-interface (SBI) that facilitates much simpler and flexible interaction amongst various control plane network functions lends itself well to the adoption of a containerized implementation.
While several established and new vendors have implemented these 5G Core network application functions as a combination of VNF/CNF on their own cloud platforms and management systems, they are continuing to evolve towards fully containerized solutions that are cloud native as quickly as possible. Some vendors and CSPs have partnered with public cloud providers to host their 5G Core applications, while others have chosen private and hybrid cloud hosted models as well.
As CSPs look to improve their revenue generation models with the delivery of new applications and services over the 5G network connectivity fabric, they may have to shift certain mobility workloads in targeted data center locations closer to the edge and/or cell sites/users. Hence the next generation of cloud native mobility core networking solutions need to be flexible in terms of their design, selection methodology, elasticity in terms of their capacity and scalability and be able to handle such workloads across distributed/regionalized and centralized data center locations in a reliable and cost-effective manner delivering superior quality of service and end user experience.
With a greater demand and potential for private 5G enterprise networks across various industry verticals, it is anticipated that more cloud native workflows could be run at the edge. To meet the dynamics of the end users/device connectivity, geographic density, mobility and end-to-end applications or service flow delivery, core networks must be flexible, intelligent, and programmable in terms of their CNF design. They need to be capable of being managed with data driven analytics and business insights that will make them promising candidates to drive innovative services in a converged wireless-wireline networking domain.
5G Radio Access Network (RAN)
Global CSPs have relied on the use of traditionally closed, monolithic vendor RAN solutions for their mobility network infrastructure deployments. The RAN is one of the most complex domains in the end-to-end mobility network that ties the end users to the core network to be able to receive their day-to-day services.
The RAN consists of cell sites or base-stations that have the radio and the baseband subsystems connected by a high-speed link used to exchange digital data. Typically, these are co-located in a monolithic cell site model, but they could be segregated so that the baseband resources can be pooled into a central location across the serving radio systems operating in multiple spectrum bands with multiple carriers.
Though standards compliant, these solutions could still be proprietary in nature to a certain extent with both radio and baseband systems delivered from a single vendor. While this has been a good practice in the industry over more than two decades, it has limited new entrants into this marketplace. With significant advances in software-defined networking and virtualization, the monolithic solutions have evolved towards virtualized RAN enabling a first level of disaggregation with a mix of specialized radio and baseband system vendors.
The O-RAN Alliance founded by a consortium of operators, vendors, research and academia has been leading the efforts to re-shape the RAN industry towards a more open, disaggregated, flexible, intelligent and interoperable mobility access network architecture design that could enable a competitive supplier ecosystem as well as new deployment models. While the O-RAN Alliance has defined various working groups that specify aspects of the architecture, white box hardware, software, subsystems and their interfaces, transport connectivity, management and control, security and testing specifications, work is still ongoing to drive new applications, features and capabilities.
With legacy 4G/3G/2G RAN equipment installed in mobile network environments and interworking required with 5G for service continuity during mobility across multi radio-access technologies, the drive towards a completely disaggregated, flexible, software defined O-RAN deployment model for 5G technology alone from a new supplier community requires considerable level of standards maturity, product design, multi-vendor interoperability, and commercial readiness proven at the ecosystem level. CSPs globally are also looking for total cost of ownership (TCO) reductions when deploying disaggregated RAN solutions that can deliver and/or exceed performance as well as quality at the cost point that is competitive to the traditional designs. Few carriers are proactively looking at evaluating and embracing cloud native 5G standalone VRAN/O-RAN solutions for private enterprise networks in greenfield environments. Although these are still in the early stages of development, considerable testing and integration needs to happen within the mobility ecosystem to understand the merits of such technology and the net value it brings to the operator community regarding operational simplicity and life cycle management.
So, what’s next with containerization in the real world?
It’s not mandatory that every part of the 5G mobility network infrastructure including radio access and core network be cloud-native from day one. The high-speed mobile transport network that glues the disaggregated RAN functions within an Open RAN environment and the 5G Core is a critical component that needs to evolve with the advances in cloud technologies and open standards. In addition, there are content delivery platforms and application servers that process significant amount of media rich information generated by a variety of digital endpoints in the mobility ecosystem that could evolve towards their cloud native implementations.
The most important aspect of software-defined disaggregated, and cloud native implementation associated with an application function (CNF) onboarding within a given networking domain is that it can carry the targeted workloads with the underlying infrastructure while delivering on its cloud native functionality, feature capabilities and measurable application as well as infrastructure network performance indicators that meet the service provider’s expectations. If they can’t deliver on the functionality, capacity, performance during various failover events, scalability, reliability and availability, then the CSPs are back to relying on manual procedures for basic tasks. This comes at the expense of coordinating operationally intensive efforts across multiple organizations, complexity and hence need to be avoided unless the CNF solutions are fully vetted out in the labs and field. Hence knowing the true performance of the CNFs in detail at the subsystem and system level and understanding their impact on the end-to-end network, applications and service delivery performance with representative real-world impairments is extremely critical to the success of domain specific application onboarding.
In the next installment in my blog series, I’ll look at how openness is enhancing network virtualization.