Telecommunications networks are undergoing radical and long-term change. It's a gradual process. But it's certain that the changes are accelerating and growing more extreme.

The reason is simple: End users want to use the networks to do far more demanding and sophisticated things than before. Consider facial identification used to identify potential terrorists at airports. The network supporting such a system must serve up enough bandwidth and be close enough to databases to perform the task between the time a person checks in and boards the plane. It must do this many times a minute.

This example involves a specific mix of speed and network architecture. There is a virtually endless list of use cases that demand various mixes of these attributes. For instance, autonomous vehicles demand far less latency than photo security: The car will need to get input instantaneously. The total amount of data exchanged, however, is less.

Manual to automatic

The good news is that ways have emerged to automate networks. Software-defined networks (SDN) and network functions virtualization (NFV) are new approaches in which the plumbing of the network and the services carried by those pipes can be changed in real time. This is a huge advance.

But the task keeps growing in complexity as well. It will take machines – in the form of network automation – to manage environments that feature the IoT and 5G. Dimitris Mavrakis, a research director at ABI Research, explained to me: “Network automation usually refers to automating manual processes that required human input and control. In the context of SDN/NFV, this means automation in the level of the SDN controller or the NFV orchestrator that can scale up/down resources according to network demands. There is no strict definition of network automation, as each network domain has different requirements, technologies and interfaces to the rest of the network. But in broad terms, it means automating manual processes.”

Closing the loop

Network automation alone won't get the job done. It must be married to artificial intelligence (AI). AI is an umbrella term that includes technologies that enable systems to change their behavior based on new data (machine learning), natural language processing (the ability to understand spoken words and, eventually, nuance), computer vision (the ability to decode images) and others.

The reality is that the sheer complexity and compressed timeframes of evolving networks make these tools necessary. According to Mavarakis: “AI and machine learning are crucial to enable automation in the telecom network. They are necessary to process the vast amounts of data networks generate and to turn these into actionable recommendations. At the moment, these processes are open-loop; they crunch data and provide recommendations to a human, who then decides whether to execute the change or not. In the long term, these AI/ML algorithms will run in a closed-loop manner, i.e. deciding and executing the commands they believe are best for the network.”

The road to AI-driven automation

Tractica principal analyst Mark Beccue offered insight into how a telecom operator should approach network automation. He suggests considering three- to five-year roadmaps that have well defined objectives and ROI projections. Some service providers (SPs) are overly ambitious, which can lead them to fall behind, he wrote. Where an SP starts from is key to its strategy. For instance, an operator heading towards 5G will have different issues than those not at the cutting edge.

Network automation is a potent tool at either the advanced or more basic levels. “SPs with advanced plans will likely be looking at node slicing/network management, as the vision for advanced networks is more real-time network slicing,” he wrote. “In this case, AI-driven automation will be impactful because it will be very difficult to manage optimally by humans. More near-term deployment use cases may look at customer experience/service assurance, using automation to predict and proactively detect network issues before customers complain or report and to automatically fix issues, reducing truck rolls for field service.”

Networks are changing with unheard of rapidity. In the battle to stay ahead, service providers should employ network automation and artificial intelligence.