By Susan White, head of strategy and portfolio marketing at Netcracker
For the world’s communications service providers (CSPs) surely nothing holds more promise than 5G network slicing. The ability to let enterprises run their own virtual private networks will open up a massive new range of lucrative use cases – in automotive, energy, entertainment, healthcare and more.
This explains why GSMA says mobile operators (MNOs) and CSPs will spend $1.1 trillion in mobile CAPEX between 2020 and 2025 (80 per cent of which will be on 5G networks). It’s a huge amount of money. But it makes sense when you consider the upside. The same GSMA study says 5G could add $2.2 trillion to the global economy by 2034.
A portion of this revenue will be the result of 5G’s network slicing capability. However, in order to truly capitalise on these potential revenue streams, CSPs must be able to manage 5G network behaviours efficiently.
5G: too many moving parts for partial automation
This will not be straightforward. Telecoms network management is already a complicated process. But network slicing just adds to the complexity, with its myriad slice attributes, multiplying relationships between elements, interdependencies and mappings.
Indeed, you could argue that slicing adds infinite configurations in terms of creating, introducing, managing and enforcing network policies. With a much larger ecosystem of players, there are simply too many moving parts for partial automation.
To efficiently manage 5G network slicing CSPs have to…
- Rapidly set up, deploy and terminate a network slice
- Ensure that network slice service level agreements (SLAs) are maintained
- Dynamically adapt to the changing needs of slice services to maintain SLAs
- Easily customise by catering to various xNF permutations, requirements and maturity levels, so they can develop new services without a significant dependence on the individual components of the underlying network.
So, what’s the answer?
I believe there is only one option for executing network policies in significant volumes: intent-based automation.
The game changer: automation based on intent
Intent-based automation refers to a business or service request, such as opening a new smart office or updating the security policy, regardless of the technical platform.
Intent-based automation is simply a game-changer. Why? Because it abstracts the service from the technical/platform implementation. With intent-based automation administrators should be able to simply express their intent, and the automated systems should carry out the task without human intervention. Everything is simplified. As a result, CSPs can adapt to changing needs much faster.
Obviously, the automation that flows from the service intent should – ideally – be total. 5G network slicing makes this more challenging because the network has to adapt to constantly changing service needs. With cloud-native network functions, frequent software releases and upgrades, private and public cloud platforms and an increase in the number of participants across the ecosystem, this automation has to be service-driven across the entire network.
But we’re not there yet. Even though full automation is the goal, we should aim to first minimise manual processes and eventually eliminate them.
Still, I believe that automating operations as described above would not only save time and cost but should also give CSPs significant competitive advantages. They should be able to exceed customer expectations by delivering improvements and launching new services quickly and efficiently.
The importance of service models
Note the word ‘should’ here. Intent-based automation is not easy to do. The quality of the automation is only as good as the service (or slice) model and how it’s implemented.
To do it well, CSPs should create a library of predefined slice service models for popular use cases. They should define the intent rules, policies, and critical parameters (e.g., SLAs and end points). And they must ensure the SLAs for each network slice are maintained – which means dynamically adapting to the various requirements of the slice services.
But here’s the thing: slices span multiple layers of a network. No single model can define all of the required operational lifecycle transitions, control-loop logic, slice-service relationships/mappings, and topology/capacity dependencies of a network slice. The best way to avoid the complexity of building multiple models is to create a composite in which parts of a data model are inserted into another.
Let’s take the example of a smart factory. In this use case, 5G connectivity, IoT applications and service test management systems all form part of the same composite service model.
How to interpret and implement the slice service model
It’s now the orchestration’s job to interpret and implement the slice service model. This important step requires a structured knowledge graph with weighted associations, which can proceed with soft decision-making.
The model can determine which transitions to make to achieve the target state. Transitions can include a modification, a scale-out or a feasibility check (or all three). Whatever the decision-making, the orchestration system needs to generate a sequence of automation steps without having to do extensive modelling for every possible state transition. And it will rely on exposed capabilities from southbound systems to implement the intent.
To return to our smart factory example, the orchestration system will communicate with the 5G core, IoT applications, edge platforms and service test management to configure the intent for the E2E 5G slice.
Execution: the power of the distributed choreography approach
This brings us to the execution stage – in other words, how the intent is executed in the orchestration system. Within a highly dynamic network, the traditional centralised workflow approach is not efficient. In fact, it becomes a bottleneck. By contrast, a new distributed choreography approach dynamically composes services in a decentralised way. This achieves greater scale and efficiency.
In a 5G world, every CSP wants to take control of the delivery of complex services and the end-to end management of network slicing. Intent-based automation – in combination with dynamic service models, new implementation processes and a fresh execution approach – will help them to do so. They will be able to simplify operations and create flexible networks that can do justice to the $2.2 trillion promise of 5G.