AI: coming to network management, for better or worse

AI: coming to network management, for better or worse

There are few skills and areas of human endeavour not already being changed using artificial intelligence, or in danger of being so, for good or ill. Network management is no exception.

In many ways it is a prime candidate for AI driven transformation. It involves analysing and acting on massive volumes of data that are already digital. In many instances, such as a network failure or cyber attack, this analysis needs to be undertaken very rapidly and appropriate remedial action determined and executed. AI has the potential to achieve these goals much faster than any human.

However, there is a downside. Adding functionality to a network inevitably increases the attack surface. Also, an AI system can be a ‘black box’ that can hide bias and weaknesses that cybercriminals will likely uncover and learn to exploit. And, as with every other discipline in which AI has made its debut, there will, inevitably, be fears that it will render skilled humans redundant.

The origins of AI date back to the 1980s but it is only in the last decade that it has really found widespread applications, and its use in network management is no exception. Several network management products developed in the 1990s claimed to use AI, and the first successful commercial product to use AI was IBM’s NetView/PC introduced in 1991.

Back in 1989 Network World reported that IBM was working to provide AI tools such as expert systems and high-level programming languages that would simplify automation of routine net control functions and ease the management of complex networks. The description sounds very close to what is claimed today for AI.

“Artificial intelligence systems will free NetView users from making decisions on the continual stream of alerts that filter up to a network management operator’s console. The systems will evaluate the alerts and take corrective measures in real time to remedy the faults or suggest a plan of action to the operator,” the report said.

A 1989 IEEE paper AI-assisted telecommunications network management described “an AI-assisted, real-time, centralised network management prototype that consists of two cooperating AI components … for the detection and isolation of communications network anomalies.”

AI has come a long way since then, and so have the complexity of networks, the challenges of managing them and the threats against them. AI will likely be essential in enabling these challenges to be overcome. And a good indication of the importance of AI in network management is the number of standards organisations working on the use of AI for network management.

Back in 2019 the International Telecommunication Union (ITU) issued a new standard (recommendation in ITU parlance) ITU Y.3172 that “established a basis for the cost-effective integration of machine learning into 5G and future networks. The standard and the use of AI in general for network management was discussed at length in a 2020 edition of ITU News.

The European Telecommunications Standards Institute’s (ETSI) Experiential Networked Intelligence Industry Specification Group (ENI ISG) has defined a cognitive network management architecture using “AI techniques and context-aware policies to adjust offered services based on changes in user needs, environmental conditions and business goals.” The group’s 2019 document ENI Definition of Categories for AI Application to Networks “define[d] various categories for the level of application of AI techniques to the management of the network, going from basic limited aspects, to the full use of AI techniques for performing network management.”

The Internet Engineering Task Force (IETF) does not use the term “artificial intelligence” in the context of its work on network management. Rather it says it is “working on standards for automated network management … to improve and make more efficient management of networks as they continue to increase in size and complexity.”

The IETF is also standardising protocols for autonomic networking with the aim of self-management of networks including self-configuration, self-optimisation, self-healing and self-protection. It says research on automated network management in its Internet Research Task Force resulted in RFC7575, Autonomic Networking: Definitions and Design Goals, and RFC7576, General Gap Analysis for Autonomic Networking and led to the creation of the IETF ANIMA working group, which finished standardising the first phase of an Autonomic Networking Infrastructure (ANI) in 2021.

So what will be the impact of AI on the roles and employment prospects for network management? AI can now provide an answer, of sorts, to that question. Ask ChatGPT “will artificial intelligence make network managers redundant?” and it will trawl the vastness of the internet in search of answers, convert its findings into perfect English and pronounce:

“Artificial intelligence (AI) has the potential to automate many tasks that are currently performed by network managers, making their work more efficient and streamlined. However, it is unlikely that AI will completely replace network managers in the near future. Instead, it is more likely that AI will augment the work of network managers, allowing them to focus on higher-level tasks and decision-making.”

So that’s the good news. However, note the qualification: “in the near future.” There is also some bad news. Ask ChatGPT “what problems will the use of AI for network management create for network managers?” And you get: complexity, integration, data quality, cybersecurity and ethical concerns. So AI looks like being very much a mixed, but inevitable, blessing.

Network engineers love AKIPS. Find out why

Get started with a free 30-day trial. You’ll receive a free demo too, before you deploy our software.