Automation for 5G Network Rollout Process

Rolling out 5G services is an opportunity for telco operators to enhance automation, improve customer experience and meet their business goals. This article outlines the route to success on each count, and shows why 5G rollout should be viewed as the start of network automation rather than its culmination.

Automation for 5G Network Rollout Process
Automation for 5G Network Rollout Process

Read now to discover:

  • The symbiotic relationship between 5G rollout and network automation
  • How intent-based models, templates and successful orchestration underpin 5G network automation
  • The role of NFV and SDN in 5G network automation
  • New approaches to on-demand and virtual network services

Content Summary

What will you learn?
Automation for 5G Ran Rollout
On-demand Services: Fulfilment and Multi-domain Orchestration in 5G Networks
NFV/SND as Enablers of 5G: On-demand Services, Edge Computing and Slicing
Summary

What will you learn?

  • Why automation is needed for 5G rollout and how to introduce it as a step by step process
  • How the 5G network itself should enable the introduction of smart algorithms
  • The role of intent-based models in automation
  • Example of 5G radio access network (RAN) rollout automation
  • Ways to implement on-demand services including network slice-based services
  • The role of the orchestrator
  • How network virtualization and software defined networks (NFV& SDN) act as 5G enablers

The main promise of 5G technology is to enable better customer experience and on-demand services, based on higher bandwidth and lower latency, over the far greater number of devices expected as a result of the continued expansion of the Internet of Things. This certainly means that 5G demands more complex networks. But it doesn’t have to mean that telecom operators become overwhelmed – provided the potential of 5G is supported with a certain degree of process automation.

While 5G specifications do define architecture enabling automation, this doesn’t mean that 5G instantly equates with a fully automated network. On the contrary, a successful 5G rollout should be understood as a process that only begins the automation journey. This might seem a dismaying concept at first sight, but scratch away the surface and it soon becomes clear that it makes perfect sense – for a successful 5G rollout should enable ongoing, constantly enhanced automation that doesn’t require costly infrastructure investments or intensive network redesign.

In fact, well-designed 5G architecture will provide the mechanism for small but important adjustments in automation algorithms, leading to a better network and improved customer services. Changes should be non-invasive, easy to introduce rapidly, and simple to withdraw if they fail to improve automation. Importantly, they should never have dramatic effects on the network as a whole.

In other words, a successful 5G rollout should give the operator the ability to experiment with new services and new automation algorithms as part of a well-controlled process. The concept of slices, which is part of 5G definition, is to help with that by enabling a dedicated network slice for experiments. Such a slice can be used to offer beta services to early adopter customers. Furthermore, having this functionality in a dedicated slice avoids disruption of already offered mature services offered in the mainstream network slice.

Automation for 5G Ran Rollout

Automation may be thought of as a process in which human intervention shifts focus from the “how” (in this case defining all details of network configuration) to the “what” (or the intent, which in our scenario means addressing customer requirements). On the network side, automation should mean the establishment of policies which lead to the generation of machine-driven strategies effective for a given operator. For the customer, automation should deliver on-demand services defined by each individual user and delivered almost instantly by the network. Going a step further, AI algorithms may provide recommendations based on behavior analysis, and suggest services which may offer perceived improvement in the quality of life of each telco subscriber.

To succeed in 5G rollout and increase the level of process automation in their organizations, operators need the right tools. 5G is a disruptive change to the telco landscape and, like all such changes, it needs to be introduced gradually. In addition, there is an assumption that 5G and non-5G networks will work together, which requires the implementation of tools capable of managing such hybrid telecom infrastructures, regardless of whether an operator’s OSS is cloud-based, locked into traditional physical hardware, or both.

At minimum, automation in the 5G rollout process should include radio access network (RAN) planning defined as high-level intent and executed in a fully automated manner. The definition of the end to end process for 5G RAN rollout is shown in Figure 1. This definition requires human input in the form of knowledge definition, which can be augmented by machine learning at the next level. Process definition may still require some manual intervention, but this should be limited specifically to intent-related tasks wherever possible, and execution should always be automatic.

Figure 1: End to End 5G RAN Rollout Process
Figure 1: End to End 5G RAN Rollout Process

A planning (intent definition) task is one manual intervention that may be triggered automatically by the end to end process. On yet another (higher) level of automation, even such a task can be assisted by algorithms based on machine learning, but it still makes sense for a human being to review the plan before acceptance. This is illustrated in Figure 2.

Figure 2: gNB Planning/Reviewing
Figure 2: gNB Planning/Reviewing

Once the planning (intent) has been defined and accepted, the end to end process automatically proceeds to the next steps. These usually involve detailed configuration of network functions. Ideally, all steps in this process should be fully automated, with the result that planning is treated as the set of input parameters to be translated to vendor-specific configurations. As achieving that targeted level of automation may be challenging, the system employs a mechanism for configuring templates, as shown in Figure 3.

Figure 3: Configuration Template Mechanism
Figure 3: Configuration Template Mechanism

The templates reveal how to configure network functions based on the planning/high-level design. However, an engineer may need to assist in the process if network functions require further vendor-specific configuration. This feeds into the next level of automation, at which example-based machine learning operates. When the same network function comes to be configured again, the system may suggest earlier parameters and, if they are appropriate, will be allowed to continue configuration automatically.

On-demand Services: Fulfilment and Multi-domain Orchestration in 5G Networks

There would be little point in a telecom operator rolling out a 5G network without being able to deliver services to match. With 5G comes the promise of revolutionary customer experience, reflected specifically in on-demand services. This assumes the next level of automation, giving the customer an impression of services which can be turned on and off as and when required.

This calls for a model-driven architecture with abstraction layers playing an essential role. At the top of the model there are customer services (customer intents), which must be fulfilled by the automation process – although technical staff must understand and be capable of creating what is required. Elements to be configured should be hidden in the lower level of the model. This top-down concept is shown in Figure 4, for a relativity simple B2B P2P Video Streaming service.

Figure 4: A Service Model in the Service Catalog
Figure 4: A Service Model in the Service Catalog

Based on this model, the customer service (intent) is decomposed by the orchestrator (see below) to technical services (low level intents), which can be requested from domain controllers by the end to end orchestrator. For such decomposition, both the model (specifications) and the abstracted view of network topology are required. Here, the abstracted view has an additional aspect. For multi-domain networks in which partners’ networks may need to be orchestrated, it must restrict network visibility to a “black box” view available only to authorized parties. Maintaining an abstracted view of domains also facilitates scalability, an aspect of particular importance for edge data centers in which a full, centralized view of the whole network will not be feasible. Rather, a cloud-native architecture designed with a distributed network inventory is required, as shown in Figure 5.

Figure 5: An End to End Abstracted View of a Multi-domain Network in Logical Inventory
Figure 5: An End to End Abstracted View of a Multi-domain Network in Logical Inventory

Based on an end to end, abstracted view of the network, the service orchestrator is responsible for delivering the service – i.e. translating the customer intent into technical intents. From the perspective of partner network orchestration, that technical intent is the high-level service for which the partner is responsible. Finally, the orchestrator creates an end to end view of the customer service based on abstraction layer elements.

This view is needed to ensure that customer service is in accordance with the SLA. A self-healing and autoscaling mechanism may need to be triggered, for which it is essential that the model of the service is divided into a hierarchy of intents augmented with polices. An end to end view of the customer service with the layered model is shown in Figure 6.

Figure 6: An End to End View of the Service on Top of Abstraction Layers for a Multi-domain Network
Figure 6: An End to End View of the Service on Top of Abstraction Layers for a Multi-domain Network

NFV/SND as Enablers of 5G: On-demand Services, Edge Computing and Slicing

NFV/SDN technologies are treated as enablers of a successful 5G rollout, and are in fact crucial from the automation perspective. Virtualization of network functions enables instantiation of virtual functions on demand, as it does not require the installation of dedicated hardware. Moreover, the same physical infrastructure can host many instances of virtual network functions (VNFs), creating opportunities for network slicing. This can be done even for enabling core network functions.

For example, separate slices for IoT and mass market mobile users can co-exist on the same infrastructure with an appropriate level of isolation. Network function virtualization (NFV) decouples the network function from its physical device, enabling a network function to be migrated to the edge of the network, dynamically and on demand. In that way, network functions can be re-allocated between network sites and adapted to customer service-related demands. Software defined networking (SDN) is a complementary technology which can be used to re-shape the network according to the re-allocation of virtualized network functions. SDN requires an end to end controller/network orchestrator (see Figure 6).

A model-based approach grounded in intent is crucial for successful NFV/SDN deployment. For example, the intent could be a request for a VNF instance with appropriate KPIs, with auto-scaling and self-healing algorithms as target values.

The modeling of VNFs should be consistent with the modeling of physical network functions (PNFs – the traditional, non-virtualized network functions). This is crucial to the success of the process, as rollouts of 5G and complementary NFV/SDN technologies are bound to be iterative, so the ability to manage “hybrid” networks is essential. An example for modeling VNFs in the Resource Catalog managing VNF specifications is shown in Figure 7.

Figure 7: Virtual Network Function (VNF) Modeling in Resource Catalog - VNF Descriptor as Resource Specification
Figure 7: Virtual Network Function (VNF) Modeling in Resource Catalog – VNF Descriptor as Resource Specification

Although instantiation of VNFs is completely different from traditional hardware-based network installations, the configuration of virtual network functions’ parameters may involve vendor-specific parameters, and mechanisms for PNF configuration may be re-used (see Figure 3).

Summary

If we are to see fulfillment of all that 5G promises, a 5G network rollout must be based on a very high level of automation. This will help CSPs to reduce operating costs and provide on-demand customer services.

Rollouts are going to be challenging, so they must be done in an iterative manner. Having the right IT tools to automate this process will be essential for operators. A successful 5G rollout is far from being the end of an automation journey. Rather, it should be marked by a greater ability to enhance automation algorithms and inject them into networks as pieces of software – similar to the processes of continuous integration and continuous delivery. This will mean that automation can be enhanced iteratively without interrupting the network’s ability to provide customer services.

While 5G and automation go hand-in-hand, telco operators need an intelligent approach to network development in order to make the most of the opportunities on offer. Our new episode explains how telcos can reap the rewards of 5G by adopting a sharply focused approach to automation in order to enhance customer experience and meet business goals.

Source: Comarch Telecoms