AIOps network management requires vendor-buyer teamwork
Before buying AIOps tools, network managers should work closely with vendors to ensure they collect the correct network data and use relevant algorithms.
Only 36% of network teams are fully effective at evaluating AIOps technology when shopping for products to enhance network management, according to research from Enterprise Management Associates Inc. Networking pros will need to develop methods and practices to evaluate this emerging technology soon, especially since 90% of them said AIOps network management can lead to better business outcomes.
EMA recently surveyed 309 network infrastructure and operations professionals who have experience applying AIOps to network management. In the report, "Revolutionizing Network Management with AIOps," EMA found a strong correlation between those who are effective at evaluating this technology and those who are most successful using it. Successful users of AIOps network management are twice as likely to say they are very effective at evaluating services.
Cutting through the hype
AIOps does have a fair amount of hype, and it begins with the name itself. Short for artificial intelligence for IT operations, AIOps is a bit of a misnomer because most of the tools don't use AI algorithms. They use a mix of machine learning, big data, predictive analytics and other advanced algorithms. This technology is good at correlating different classes of data and identifying patterns and anomalies.
With proper training, more advanced AIOps services can draw conclusions based on the patterns and behavior they observe. Then those tools can act on those conclusions, either by presenting insights to IT personnel or by taking automatic action on the network. Some vendors might stretch beyond machine learning and start to use AI algorithms, but for now that's quite rare. So, expectations should be tempered.
AIOps isn't going to replace humans in IT operations. Instead, network management and infrastructure vendors are building AIOps that can simplify their technology by automating certain tasks, such as anomaly detection, alerts and escalations, and root cause analysis.
In the meantime, the AIOps hype can make it hard for an IT organization to evaluate the technology. What is AIOps? How does AIOps work? And which AIOps software is right for me? These questions can be tough to answer.
As an example, a network architect with a $40 billion pharmaceutical company told EMA that he is frustrated by the webinars and digital marketing efforts that vendors champion around AIOps.
"Everyone uses the same buzzwords," he said. "It's the same as when SD-WAN came out. Everyone was talking about it, and no one knew what it was doing. Give me a 10-minute presentation that tells me what you have achieved, what real-world problems you have solved, even if it's just in a lab. Show me what you have analyzed and what the outcome was."
Evaluating AIOps for your network
EMA asked research participants to share their preferred methods for finding the right AIOps service. Fifty-five percent said it's all about understanding the data that vendors use to train their algorithms. They want to know if vendors are using the appropriate volume and variety of data to ensure AIOps technology fully learns how a network is meant to behave.
In many cases, the vendor will need to train its software on the individual customer's network. Thus, a buyer needs to work with the vendor to ensure the tool is fed the right variety and volume of data from the buyer's own environment. Network managers cannot skip this step. Nearly 20% of research respondents said they struggle significantly with AIOps tools that are not sufficiently trained on their network.
Another key part of evaluating AIOps technology for networking is a review of the algorithms the vendor uses. Buyers don't have to audit a vendor's algorithms. Chances are a vendor won't allow that anyway. Instead, buyers should study the types of algorithms that are useful for AIOps network management. Then, they can talk with prospective vendors about the choices they make in developing such algorithms. Fifty-three percent of research participants said this review is important to AIOps product evaluation.
Finally, network teams should follow their tried-and-true processes for product procurement. Overall, 49% of network teams lean heavily on vendor resources like data sheets, case studies and product demos. Consuming vendor resources is a best practice for evaluating AIOps, according to EMA data. Another reliable approach is a proof-of-concept implementation; 46% of network managers said this step is essential.
AIOps is still an emerging technology, but EMA believes that network infrastructure and operations teams should be evaluating this technology now. Most network management and infrastructure vendors are offering at least rudimentary AIOps capabilities today, and many more will be introducing services by next year.
In 2022, AI network management is all about growing trust