‘Network slicing is a necessity in the context of 5G’ – Layer123 talk to Nokia's Kurt Pynaert

In the lead-up to the Layer123 Network Slicing Webinar, Layer123’s Millad Farahani caught up with speaker Kurt Pynaert, the head of Nokia’s end-to-end SDAN solutions team, to talk about what network slicing can make possible and the characteristics to bear in mind when provisioning slices – and it’s much more than a question of bandwidth.

 

With the rollout and uptake of 5G across a vast number of industries, we're seeing the crucial role of network slicing in empowering service providers to offer innovative services and expand into an array of new markets. What is your view on network slicing and what is it about in relation to 5G rollout?

Let's start with explaining what network slicing means in the 5G area. Network slicing is about sharing a physical network in the most economical and effective way amongst different tenants. What makes it so specific for 5G is that 5G comes with a new set of services with diverse requirements, so it's not only about speed and bandwidth – these new services require ultra-low latency and ultra-high reliability as well. To transport those new 5G services over the network in the slicing context, each slice needs to be configured, provisioned, assured and optimised for a specific set of requirements, which are quite diverse amongst the different services. This is not only about providing mobile services, but also about using 5G and network slicing for industrial automation, industry 4.0, fixed-access wireless, and so on.

In summary, network slicing is about utilising the network for the specific services that come along with 5G. We could do this in the same way we did in the 4G era, but with 5G one size doesn’t fit all. In the pre-5G era we transported low bandwidth traffic, high bandwidth gaming, metering and video streaming within the same sort of pipe. But in the 5G era, because these all have specific requirements, we slice the network so that each service has its own slice in the network with specific characteristics.

What is important as well is that things like service level agreements are catered for when providing a sliced network in the 5G era. Service level agreement is not just about bandwidth and speed, but also latency. If there's a low latency service that the network or slice is devised for, say for instance an application of e-health using network slicing, there is no need to explain that this connection should be highly reliable.

Network slicing in the context of 5G automation is a necessity. It's essential because you cannot provide such different characteristics by creating network slices manually. This means automation is key to network slicing.


Could you go into more detail on the different characteristics that you defined within network slicing?

Among the different characteristics we are all used to is speed - the speed of the Internet, the speed of downloading or streaming videos. This is still of course a key characteristic, but it's not just about speed and bandwidth. If we look at applications of e-health, industry 4.0, automated industries or smart cars, latency is much more important than speed. A service can have very low speed but very low latency as well. This means latency is a very important characteristic in the context of network slicing and 5G.

Alongside speed, bandwidth and latency, there is also:

Reliability - if you are using the network to set up an end-to-end slice, for instance for an e-health application, the communication and the slice need to be very reliable. This also means you are not just setting up the network slice once, but continuously monitoring the characteristics of the slice to make sure that the service is being delivered according to service level agreements.

  • Security – the moment you start sharing a physical network among different tenants used for different services, security becomes key. Isolation and security over the slice are very important in this context. Additionally, it's not only about providing traditional services like voice or high-speed Internet, but also things like analytics, data analytics, augmented intelligence, and closed-loop automation.
  • Coverage. It could be that a network slice is mobile - for instance the endpoint connected to the network slice using the service at the other side of the slice is not static, but mobile. If you move from location A to B, it is important that the slice is designed for that. This implies both coverage and mobility. For instance, if you would use network slicing in a 5G context with drones - needless to say, drones are mobile. In general, not only is end user equipment dynamic and mobile, but we also see the density of end of end user devices growing. I call this the densification of the network. It’s not about your terminal or PC at home, but about millions of different devices connected on that network using those network slices. So device density is key as well.

And then there is the whole shift to cloud native-controlled applications, where you are not only fulfilling the service but also controlling it –assurance, monitoring, health and network optimisation in order to offer the expected service at the lowest possible cost. This is the digital future of the network – having individual network slices for new requirements in the context of emerging 5G services.

 

What challenges do you see within end-to-end orchestration and management of network slices?

Automation is a necessity in in this context. In an end-to-end context, you need to connect different parts of the network. So you have the access network, the transport network, the core network – all those network domains need to be connected.

But all these domains have some specific characteristics. The first thing you need to automate is to create a network slice as a virtual overlay over those different domains. This means you need to have knowledge of the domains, but you also need to delegate the specifics of the domain to the domain controllers.

This is overarched by what we call in Nokia an orchestration function, where you orchestrate based on a request to create a network slice and a NaaS, to orchestrate amongst the different domains in the network. This is not just setting up the connection, but also making sure the whole slice is monitored and assured so that the data is analysed and SLAs for each segment are guaranteed, fine-tuned, and optimized. This is all required together with the move to cloud controllers – the need to connect, orchestrate, manage, and control the end-to-end network slice. If you sum this all up, it means you automate the whole solution.

What we are working also implies a different approach to setting up an end-to-end slice. This starts from a business intent – what we want the network slice to achieve at a business level, which is then broken down into the different domains. You need to design this kind of orchestrated controlled network slice and then deploy it, and once it's deployed you need to assure that the slice is delivering its expected service level agreement.

Another complexity is that the multi-domain and multi-vendor nature. An end-to-end network is not just built by one vendor, it is a mix of different vendors and parties, both hardware and software. So as part of this automation it is important to connect and orchestrate the domains in a vendor-agnostic way. And these are the solutions we are working on.

In summary, it's about translating the service request defined according to the characteristics I stated before (latency, throughput, reliability, mobility, geography, and so on), putting all the composable network elements together, connecting them, orchestrating them, and delivering the service according to the design across the different domains the network is built with.

 

Kurt Pynaert will share more of his network slicing knowledge as part of the upcoming Layer123 Network Slicing webinar, alongside speakers from Bell Canada, Microsoft Azure, Orange, SES Satellites and Amdocs. Register for free here if you’re a network operator or service provider.