BT's McRae on Taking 5G to the Edge

Investments at the 5G edge will increase massively as the performance of the mobile network is brought up to par with the fixed-line networks, according to BT's Neil McRae.

The Holy Grail on the path to 5G, or any other new deployment, is to simplify platforms and systems to make the large investments profitable. In part one of this Telco Transformation Q&A, McRae, managing director, architecture and technology, and chief architect at BT , spoke about the challenges and solutions for 5G deployments.

Telco Transformation: BT has vocally questioned 5G's financial viability on the grounds of the paucity of use cases, and the high investment costs despite the expected lowering of hardware costs with virtualization. Can you talk about the challenges of 5G that impinge on its profitability?

Neil McRae: 5G NR [new radio], whose standards were approved recently, is a whole new hardware deployment, except for some 4G elements like MIMO and SON. It will increase capital expenditures owing to the overlap of moving from 4G to 5G. To exploit the capabilities of 5G NR -- covering a wide range of frequency bands such as 6GHz and millimeter wave spectrum -- we have to overbuild the radio hardware and even antennas, which is a significant task. We will need to invest enough on the 5G infrastructure so that customers buy 5G devices and services while simultaneously maintaining the 4G infrastructure. Many of our partners are building platforms that mitigate some of the risk of 5G investments, but managing these investments to get the most out of them is a non-trivial challenge.

While 5G will move at an accelerated pace towards virtualization and SDN controls, which consequently save on hardware costs, we are concurrently building a distributed network that will need investment to bring converged infrastructure resources closer to customers to deliver the 5G experience.

TT: So how big of an investment, and what else is needed?

NM: To give a sense of the magnitude of the investment here in the UK, the fixed network today supports peak traffic of 10 terabits per second compared to the 350-400 gigabits per second on the mobile counterpart. To scale the service volumes and quality of service at the edge, mostly in large metros, we will need a different approach.

Virtualization software and SDN controls save as many costs as they add with licensing royalties and complexity. Open source is an alternative to licensing, but it needs development effort, which can increase costs. Virtualization technology helps with balancing capacity but the network is always so busy that benefits are limited in this area.

Orchestration is a critical benefit but even after several years it is still far from delivering the promise, which shouldn't be a surprise given the investment operators have made in OSS over the years. It is hard to build a platform that meets all the needs for network management. The subsystems of the platforms are not vendor-agnostic -- the suppliers add their management and orchestration software that are not necessarily the most optimal choices.

SDN is a different story and it will help massively -- it is going to enable us to automate operations. The telco industry's effort to automate with OSS has not been a great success historically. SDN allows us to integrate at a much lower level with the network infrastructure, and enable a proper closed-loop architecture where you do something, monitor it, make changes and start over. Tied to that are service-based models that allow us to define services clearly enabling scalable automation and the use of telemetry, reducing the cost to run the network and getting much better yield on equipment deployment.

TT: One of BT's studies (registration required) indicates that network slicing can be profitable with the number of services as few as five, and a payback period of just three years. Does that mean we'll see a rapid adoption of network slicing?

NM: Network slicing will be absolutely beneficial in 5G networks with their heterogeneous use cases, devices and applications. I believe that network slicing is a first step to something more; towards microservices and a distributed architecture-based means to manage all the different use cases and -- more importantly -- be able to scale them.

Network slicing adds a lot of complexity and drives increasing importance on understanding dependencies -- it's like managing several networks. Network operators have to monitor to see if they are all working as expected and ensure that they are delivering on their performance, bandwidth and capacity. There are tools to cope with the complexity, but they are in their early stages of development; I would argue the success of network slicing depends on how well these tools develop.

TT: So if there were one thing you could ask for today what would that be?

NM: We need simplicity in the tooling. I look at BT and see very complex platforms and systems that are hard to manage. Any new technology deployment must drive simplicity and automation. Our approach is going to be to pick-and-choose components, instead of whole systems from some vendors, to achieve radical simplicity where we integrate this technology, ensuring that we manage the architectural dependencies and allow us to maximize the benefits of slicing and automation. We will have specialized vendors in individual domains like SON, AI, and low latency, but we don't believe one size fits all.

Related posts:

• Networks Need Telemetry – BT's McRae
• BT's McRae: SDN Is All About Automation
• BT's McRae: Common Service Models Aid Automation
• Climb the Highest Mountain: BT's McRae

— Kishore Jethanandani, Contributing Writer, Telco Transformation