What are the applications of machine learning in network management?

As networks move toward automation and intelligence, enterprises are increasingly demanding artificial intelligence (AI) and machine learning (ML) because they can programmatically identify network issues and perform instant diagnosis of complex problems.

Applying AI and ML to network management enables the integration of inputs from multiple management platforms for centralized analysis. Rather than having IT staff manually comb through reports from different devices and applications, machine learning can quickly and automatically diagnose problems.

“I have a lot of monitoring tools, and they all tell me something is wrong, but they don’t tell me where,” explains Josh Chessman, senior director and analyst at Gartner. “The biggest advantage of machine learning is that it can specifically identify 26 network issues across seven different tools.”

Analysts say enterprise adoption of such monitoring tools is still in its early stages. One sticking point is what do AI and ML really mean? Those who imagine AI as being able to effortlessly identify intruders and analyze and optimize traffic will be disappointed.

“Using the word AI to describe what is actually happening with new network management tools is an overstatement,” said Mark Leary, research director at IDC. “When vendors talk about their AI/ML capabilities, if you take an honest look at it, they are talking about machine learning, not artificial intelligence.”

There is no strict definition between the two terms. Broadly speaking, they both describe the same concept — an algorithm that can read data from multiple sources and adjust its output accordingly.

According to experts, artificial intelligence is more accurately applied to reliable representations of this idea than systems used to identify the source of specific problems in corporate networks.

Jagjeet Gill, a leader in Deloitte's strategy practice, said: "We may be over-interpreting the term AI because some of these things, like predictive maintenance, have been in the space for a while."

Another sticking point is cross-compatibility.

Right now, most of the products on the market are in the form of vendors adding new capabilities to their existing products. For example, many vendors are adding AIops because it's a bit of a buzzword.

There are also some vendors that can use machine learning to enable enterprises to apply AI operations and focus on IT event management, such as Moogsoft and BigPanda. But it is more common to bundle ML capabilities with products from specific vendors.

Regardless of the hurdles this technology needs to overcome, ML technology will likely make life easier for many IT professionals. Having these types of tools and solutions is a good thing and will help you stay informed about everything that’s happening in your network.

While this could be a significant step toward full network automation, it could also result in job losses for IT staff.

More likely, machine learning will free up IT staff to do more revenue-generating activities rather than put out fires. Full automation is still a long way off.