• Fulfilling the Promise of 5G – 5G requires a fully automated and intelligent network infrastructure capable of delivering high speed broadband and new enterprise and industrial network services.
• Taming Complexity – AI and agile cloud-based resources are necessary to deliver a fully digitized and flexible services environment, simplify complex network management and reduce operational costs.
5G network expectations have been well articulated, but meeting these lofty expectations is another matter. Today’s networks are multi-layer, multi-technology, and multi-vendor, which adds to the depth of the challenge at hand. The goal is to deliver new network and application services in a manner that satisfies on-demand user expectations over a network infrastructure that grows more complex over time.
The good news is that automation and AI technologies are steadily ramping up to address the complexity inherent in these multi-X networks; however, before automation and AI can be instrumented, the network infrastructure must be made programmable, a monumental task. The following analysis explores each of the key capabilities needed to establish an intelligent 5G network and note significant advances in vendor solutions.
Programmable network elements are now prevalent in the IP, transport, and mobile networking domains, and are beginning to be implemented in the fixed access domain. The IP domain has been simplified through the use of protocols such as segment routing (SRv6), Ethernet VPN (EVPN), and others. It has taken SDN years to mature, but it now forms the foundation of most IP networks.
• Cloud Resources: The use of cloud-based resources has become engrained in service provider and enterprise networks. 5G networks need to support a diversity of services and provide the ability to seamlessly support private, public and hybrid cloud deployment models. As new 5G use cases evolve, such as IoT and Industry 4.0, cloud usage will increase. Cloud providers offer service models that can support traditional compute and storage in support of low-latency services based at the network edge. The network operator needs to be able configure cloud services as seamlessly as it does network resources from the private IP domain, for example. (GlobalData will explore the business or commercial challenges related to cloud resource support in a future commentary.)
The adoption of cloud-based services has also grown in popularity and usage, with operators now being able to place workloads based on customer preferences; major cloud providers offer services via APIs to make the migration process nearly seamless. All major cloud providers now offer telco services; as a result, vendor/operator solutions need to be cloud-agnostic and able to support billing and service agreements between users and providers.
•Automation and AI: Automation and intelligence are critical capabilities. To facilitate their adoption, multiple vendors and operators now support the TM Forum’s Autonomous Network framework (AN) which consists of three layers (resource operations, service operations, and business operations layers) and four closed loops (user, business, service, and resource). The framework addresses single-domain autonomy and inter-domain collaboration, enabling a pragmatic approach that takes into account various telco operators’ unique priorities and the reality of a long coexistence of legacy and new technology based networks. The model also defined six levels for AN, ranging from “manual management (L0)” to “full autonomous network (L5)”, with most of currently deployed networks falling between layer L2 (partial automation) and L3 (conditional automation). In addition, standards and industry organizations such as ETSI and 3GPP have also contributed to AN progress, and through the joint efforts of all industry players, AN is becoming a reality.
AI comes in to play at all levels, but at Level 3 it begins to deliver significant advantages in the operation of the network by providing analysis and decision-making capabilities to assist in O&M and delivering on customer intent.
Source: TM Forum Autonomous Networks- whitepaper
Automated functions, in current solutions, have evolved to reduce the need for manual configuration of network resources and services. Levels 4 and 5 rely on advanced AI to deliver a fully autonomous network, and are several years away from fully maturing. Vendors, such as Huawei, that have been strong proponents of automation, note they are targeting 2030 to satisfy the Level 5 maturity criteria of full automation. However solutions at that point may not be fully mature from a commercial/customer deployment perspective, since it may still take some time to complete the implementation on their networks.
• Solution Architecture: The following network graphic provides an architectural view of an end-to-end 5G network. At the foundation layer (i.e., the domains) there are multiple functions needed to manage and control that domain. Current solutions now include the ability to apply Policy (SLAs), Machine Learning (ML), Telemetry (insights), Intelligence (AI), and Security within each domain and communicate that to higher layers of the network, such as the resource orchestration layer as shown in the graphic.
AI is included in the resource and service orchestration layers to provide insight into the state of the network resources and services, respectively. Each network is different and in some cases the capabilities of a specific domain (e.g., IP) are adequate to effectively manage the network resources and support the required services. However in the case of a large multi-domain CSP network (IP, optical, mobile, etc.) the domain controllers will interwork in a hierarchical model to manage inter-domain policies and provide coordinated inter-domain control.
Current solutions have evolved and provide a level of automation and AI within each domain; however, correlation in multi-domain/multi-layer networks is being performed at a level above the domain controllers, usually the resource orchestration layer.
• Solution Landscape: Multiple vendors are developing and operators are fielding solutions which follow the TM Forum guidelines for autonomous networks. AI is being enhanced and used to improve service delivery and support greater levels of automation into the automation cycle to truly realize zero-touch experience of the service lifecycle operations management and conquer the complexity issues noted above. As more advanced 5G services begin to be deployed, AI and automation will continue to advance to provide a greater degree of intelligence. A combination of tools including robotic process automation (RPA), intelligent business management software (iBPMS), and AI are being developed. These along with other puzzle pieces – including (modernized workforce, AIOps, and IT/network operation alignment, will all contribute to the goal advancing the industry from its current level 2-3 stage for most operators toward eventual full network automation.
Although significant progress has been made, continued collaboration between vendors, operators, standards organizations/forums and end users needs to continue to address the new use cases which bring forth new challenges. The industry is undergoing a multi-year transition toward autonomous networks, which in turn is fundamentally changing network architectures. GlobalData has separated solutions into two layers – resource orchestration and service orchestration – to characterize the automation and AI functions needed within each network layer. The long term goal is to minimize human involvement in lower level network functions and leverage automation and AI to deliver on network intent.
Summary: While AN has progressed, it will take years before automation practices and capabilities are in full swing. However, beginning in 2021 new private network use cases are emerging rapidly which will require networks to advance beyond level 3 in order to cope with complexity and execute intent-based decisions related to network performance and quality of service.