IoT Edge Computing / Machine Learning / Networking

Kentik Turns AIOps Spotlight on Network Data, Workflows

29 Jul 2019 9:51am, by

San Francisco-based startup Kentik, which has focused on real-time data for network traffic intelligence, is jumping on the bandwagon for artificial intelligence-aided operations (AIOps), touting capabilities specifically for network professionals.

“If you look at the event correlators or pure-play metrics or log vendors, they have not focused on the network context,” said Avi Freedman, Kentik CEO. “They may say there’s a problem over in the network, but what is it? …We’re embracing [the network], but taking a more AI approach to surfacing insights and automation approach to what you do with that.”

The AI-enabled capabilities include:

  • Network operations insight into infrastructure and traffic across cloud, data center, WAN and campus environments, including traffic growth and capacity run-out dates.
  • Edge network utilization and costs, including predicting cost overages and alerting on traffic spikes so teams can shift traffic to avoid network congestion.
  • Network protection by setting smart baselines and thresholds to automatically recognize traffic anomalies, more easily investigate incidents such as DDoS attacks, and automatically prevent threats from causing performance and availability issues.

The majority of Kentik’s early customers are service providers. AIOps can help them understand how their customers and subscribers use their services to more quickly discover service issues, usage patterns that affect customer experience, and their costs per customer, according to the company.

Taking a Big Data approach, the five-year-old company built a data engine modeled after Google Dremel to ingest data from multiple sources, then enrich the data with things like application context and routing information.

“We were solving the data problem and not so much the workflow problem,” Freedman said of its early days.

“… [Now we’re] continuing the evolution to say we want to give the practitioner more proactive notifications so they have to do less — we’ll call it spelunking — and have built-in workflows … We’re doing more integrations — integrations with orchestration systems, being able to push some of these things that are more human-driven toward more RPA [Robotics Process Automation] and closed-loop automation.

“Closed-loop automation is not a reality for most customers,” he said. “How do we take the things that are trusted insights and automate those? …We’re trying to make it so you don’t have to be the expert to get value from the infrastructure.”

AIOps, considered one of the most overused buzzwords in the industry, is getting plenty of attention as a means to automate tedious, routine tasks.

IDC has predicted that 70% of CIOs will aggressively apply data and AI to IT operations, tools, and processes by 2021. Network management is considered a prime area for applying machine learning and artificial intelligence, though a greater percentage of enterprises are looking into it than actually have it already in place, according to Gartner.

Gartner and Forrester have included a lot of vendors that do a lot of different things in their take on AIOps, Freedman said.

“…Our take on AIOps is that when we surface these insights, not only do we drive workflows, but those workflows interlink with other layers of the platform,” he said.

Going forward, the company will be focused on more integrations, workflows and automation.

“Our monitoring has been very ‘what is’-focused: What is your traffic? What is your application? What is your infrastructure? We’re going to be adding in synthetic monitoring from a network perspective,” he said of adding more “what-if” scenarios.

He cites ThousandEyes as a leader in this type of monitoring. Meanwhile, Palo Alto, California-based startup Nyansa also is taking an AIOps approach to the network, though focused on user experience at the endpoint.

Kentik’s AIOps platform is available now for existing customers. The platform will be generally available in October 2019.

Adoption figures for AI, ML, and intelligent automation are all over the map depending on how the terms are defined and who is asked about it, The New Stack analyst Lawrence Hecht reported recently.

The New Stack contributor Janakiram MSV, a principal analyst with Janakiram & Associates, and Steve Burton, vice president of marketing at, discussed how AIOps is taking DevOps to the next level in an episode of The New Stack Analysts.

Feature image via Pixabay