Ninety-two percent of IT organizations are interested in using AIOps or advanced analytics features embedded in or native to their network performance management tools, according to Enterprise Management Associates research.
AIOps is shorthand for artificial intelligence for IT operations. The definition is flexible, but Enterprise Management Associates (EMA) generally considers AIOps to be a technology that uses machine learning algorithms and other advanced heuristics to enhance the capabilities of IT operations management tools.
The use of the term AI is a bit of a misnomer. AIOps makes IT tools smarter and more effective to perform certain tasks like event correlation and automated root cause analysis. But the tools are not necessarily intelligent. In fact, EMA does not expect AIOps to take humans out of the IT operations equation any time soon.
While several stand-alone AIOps vendors process data streamed from IT management tools, many management tool vendors are developing their own AIOps technologies to enhance the value of their own products.
AIOps required in network performance management
EMA recently published research titled "Network Performance Management for Today's Digital Enterprise" based on a survey of 250 IT professionals and one-on-one interviews with subject matter experts. The survey found 60% of enterprises are currently using AIOps features embedded in their network performance management tools, and 28% of them said these AIOps features are critical to their organization.
Another 33% said they plan to adopt embedded AIOps features in their network performance management tools in the future.
"We're just starting to use it," a managing director of infrastructure at a large North American financial enterprise told EMA. "One challenge is how to do the analysis. Our [network performance management] vendor has a public cloud analysis engine, and we're very sensitive to data exposure. On the other hand, in-house analysis can be cost-prohibitive."
AIOps use cases for network performance management
EMA's research explored how enterprises identify their most important use cases for applying embedded AIOps features in network performance management tools.
The top response, selected by 28% of survey participants, was automated network traffic analysis. Specifically, this involves the ability to detect and identify traffic patterns, such as anomaly detection. This is actually the first use case offered by some early entrants in this area. For instance, ExtraHop Networks' cloud-based machine learning service Addy initially came to market as an anomaly detection feature.
Three other use cases tied at 26% each as top secondary AIOps use cases.
First, enterprises are interested in automated root cause analysis of network problems. Next, they are seeking automated capacity management, such as the ability for a network performance management tool to suggest an action to avoid future congestion. Finally, network performance management users want automated security remediation, such as suggested or automated actions in response to an incident.
Several network performance management vendors support many of these top AIOps use cases, such as Accedian, Cisco and Nyansa.
Is AIOps easy to use?
The machine learning algorithms in most AIOps services take some time to learn an environment before they deliver full value. One can analogize this to a newly hired network engineer learning the ins and outs of an organization's network.
However, enterprises should make sure their vendors can facilitate the onboarding of AIOps features and the ongoing care and feeding of these capabilities. There is a lot of complex computer science going on behind the scenes, and enterprises need to be clear upfront about how much of that will be handled by the vendor.
In fact, 42% of enterprises said they have in-house data science experts to support some AIOps technology. Another 34% said they engage with their network performance management vendor's professional services to support some data science requirements.
"In my mind, data scientists are what's missing from engagements [with AIOps features]," a senior consultant with a North American IT operations consultancy told EMA. "I had to become a data scientist. We were having an issue with an AI algorithm, and no one could figure it out."
Regardless of ongoing support issues, enterprises clearly recognize the value of AIOps use cases in network performance management products. EMA recommended that IT buyers ask their network performance management vendors what their AIOps roadmap is. If a vendor doesn't already have advanced analytics features, it should at least have plans for how to address customer requirements in this area.