This article was originally published on Tractica’s sister site Light Reading.
Telecom service providers will use AI to manage and operate networks or many of their businesses won’t survive.
That’s one of the key findings of our recent report, and it’s based on the simple economics of price, cost and profitability.
Telecom is a capital-intensive business with high fixed costs, which puts pressure on service providers to control variable costs, especially human capital. This has always been an issue, but recently it’s getting worse.
In 2017, Tom Nolle of CIMI Corporation estimated that many CSPs crossed the point where revenue per bit was lower than cost per bit in 2017. Threatened by fast and highly efficient web-scale companies, service providers are straining under the challenge posed by digital transformation. On top of that, they must solve how to manage and operate the dizzyingly complex 5G networks profitably. IoT use cases, ranging from autonomous cars and industry sensors to smart cities create a multiplier effect for network complexity. Most experts believe 5G network monitoring and management will largely be beyond human control, and that AI-fueled automation is the only way CSPs will be able to offer future network services.
While it is important to acknowledge that many automation processes for networks will be deterministic, telecom is a business ripe for AI, which dangles the promise of lowering costs and boosting efficiencies, and the biggest target is network management, and nirvana is a fully (or near fully) autonomous network.
So, how are things looking?
In our report, Tractica forecasts spending on AI-driven network management software will grow from $23 million in 2018 to more than $1.9 billion in 2021. In relative terms for service provider capex and opex spending, that’s not a lot, but we believe 2021 will be the inflection point, as telecom operators increasingly adopt SDN and NFV and begin to deploy 5G networks and leverage comprehensive network automation solutions. As telecom operators shift their capex investments from hardware-centric networks to software-based, open-source network opex investments, AI-driven network operations will command an increasingly larger proportion of overall network investments.
By 2025, Tractica forecasts annual spend for AI network operations solutions to reach $7.4 billion.
However, there are significant challenges for AI network operations solutions adoption: the adoption of SDN and NFV, struggles in sufficient cross-vendor (AI) interoperability (read: vendor resistance to open source) and an ongoing demand for deterministic automation solutions.
We delve into all of these issues in detail in the report, but for now, let’s take a look at an example of a practical path to increased AI network management:
Networking software player Aria Networks has built an AI-based network optimization solution that, today, goes beyond prompting network engineers. Its customers include Verizon, Telus, Level 3 and other Tier 1 operators.
“If the goal is automation (and perhaps even autonomy), it’s an incomplete strategy to focus only on identifying patterns,” wrote Robert Curran, CMO, in an email to Tractica. “In a network-centric business like telecom, understanding how decisions are made about how, and when to change the network are the real heart of the operation. That’s where automation efforts should be focused.”
Curran gave some examples of where its automated network management solution is solving network issues:
One of our customers has to send out IPTV signal across the network. For protection purposes, it has to be sent from two different studios. The problem is how do you compute that so the service meshes and syncs? You need to do it so you can optimize and get more traffic on the networks. It’s a timescale complexity issue. Using conventional methods, you would have to pre-test it a week before. Another example of how the solution works might be maximizing cost efficiencies. A video service sending content to a location, the cost of that from a data center, power and transport costs vary, so if you could move that to a more cost-friendly center, that would be good. A third example would be ensuring revenue streams. Let’s say you are televising a Formula One event and want to sell click through. If that signal degrades, you lose revenue, because you can’t sell that click through.
The road to network automation is an uncertain one, but the market will learn, as it always does, while pioneers take risks to blaze the trail.