Follow Datanami:
October 2, 2019

Lack of Visibility in AIOps Concerning, BigPanda Survey Says


Nearly one-third of the folks who took a recent BigPanda survey said they are evaluating AIOps tools to help them get a firmer grip on increasingly complex IT operations. However, the lack of visibility into today’s AIOps tools is a real concern, according to the survey.

BigPanda yesterday released “The Future of Monitoring and AIOps,” a study based on a survey of 1,300 people in IT operations (IT Ops), network operations center (NOC), and DevOps professionals. The big takeaway form the survey is no surprise: IT pros are under pressure to get more done at their jobs.

Specifically, roughly 50% of the IT pros surveyed cited the heightened pace of code changes as a concern. Other concerns included the accelerating rate of cloud migration and the overall increase in workloads.

IT professionals rely on an array of tools to get work done. But lately, it appears the toolshed has been growing, as 42% of respondents said they used 10 or more tools to get their jobs done.

It’s no wonder that folks are looking to AIOps to help simplify the situation. In fact, some experts say humans have no hope of monitoring today’s sprawling, containerized digital properties that are generating log data — much less managing it — without the power of ML. You can count BigPanda, which develops AIOps tools, in that category.

About half of BigPanda survey respondents say they feel more automation, including in the form of ML and AI, will help by giving them more control over their workloads. A smaller number (one-third) are actively evaluating and researching AIOps tools at the moment.

But according to BigPanda’s survey, AIOps may not be the panacea it’s made out to be. The company’s survey found that 79% of respondents want to see the ML logic behind the AIOps tool, 81% want to be able to edit the logic, and 85% want to preview results before deploying the ML logic to production.

“Thus far, the majority of AIOps tools have not placed the same emphasis on incorporating transparent, ‘explainable AI’ into their tools,” says Elik Eizenberg, CTO and co-founder at BigPanda. “This is a problem because the lack of transparency in enterprise IT tools, including AIOps tools, is a barrier to adoption and usage.”

Eizenberg says human IT Ops, NOC, and DevOps teams “can’t make decisions based on insight from AI if they can’t understand and trust it – especially when those decisions might result in an outage or prolong an outage.” With downtime costing upwards of $5,600 per minute (a figure he got from Gartner), the stakes are high, he says.

BigPanda has been a pioneer in adopting explainable AI into its software. “With BigPanda’s Open Box Machine Learning, teams can see the ML logic in plain English, they can edit the logic, test it and run what-if experiments before deploying it to production, and they can add situational and tribal knowledge that strengthens this logic,” he tells Datanami.

ML and AI is inherently complicated, as it’s based on huge matrices of long numbers, Eizenberg says. “This barrier can certainly be overcome but it’s not easy and takes years,” he says.

Related Items:

Can AIOps Save IT Management?

Data Science Helps Troubleshoot the Datacenter