Rethinking AIOps – It’s all about the data

I recently sat down with Rick Fulwiler, Senior Director of Product Management – AIOps at NETSCOUT, for an in-depth and insightful discussion on the foundational role that good data has in ensuring that service provider AIOps delivers on its promise. You can watch the entire discussion here, but the following are some of the highlights.
In network operations, we are always looking for areas for improvement.

• How can we manage applications to best serve end customers?
• How can we use end-to-end visibility of application and service performance to improve customer experience?
• Will anomaly detection for automated pattern discovery and prediction improve the quality of experience?
• What is the best way to use software-driven networks to support new services and monetization strategies?

AIOps can help to address these questions and drive transformation in network operations across all lines of business—IT, enterprise, cable, security, and wireless. “CSPs are looking at automation technologies to cut costs and boost productivity. AIOps provides a new competitive edge that the industry needs to tackle some of the automation challenges and can drive untapped data monetization revenue streams,” Rick said.

Rick thinks that AIOps is more than a buzzword. “AIOps use cases include improvement of network operations efficiency in the RAN, MEC and core; enhancement of customer experience; insights into customer behavior and preferences; improvement of Net Promoter Scores (NPS); identification of unique traffic patterns and anomalies (e.g., heavy users on FWA who cause spectrum exhaustion via hosting or illegal usage sharing); driving data monetization; network slicing assurance for enterprise customers; and roadway traffic analytics for usage and pattern prediction and planning.”

But we need to avoid the risk of adopting AIOps hoping that it will compensate for the lack of reliable, relevant, and curated data.

Rick explained, “At NETSCOUT, we call network data the hidden gold, which must be mined and refined. We all become enamored with the cool AI parts of AIOps, but we must avoid losing sight of the critical data that drives the AIOps engines.  In 5G standalone networks, the hidden gold needs not only to be mined but also refined to become consumable by the AIOps pipeline. That is critical. We don’t just take network data as is. We add context and meaning to it and define the relationship to other data being collected. We are not throwing packet-level data in the pipeline or into a data lake and expect magic to happen.”
“Identifying the relevant data and providing context for selected use cases allows us to exclude what is, effectively, noise - and causes AI models to become bloated, slow, and power inefficient. AIOps does not need all the packet-level information; it only needs the golden nuggets in those packets that are relevant to solving problems. The secret sauce is understanding what data is relevant and correlating the data required to solve specific use cases,” Rick stated.

Is AIOps going to be introduced in one go? Rick believes we need to be patient. “AIOps sets up a nice crawl, walk, and run strategy that allows carriers to go through an adoption period to learn how to leverage the derived statistical network resource usage data the AIOps calculates to improve the service operation efficiency of the 5G SA network. This adoption period can have multiple phases as carriers tune the AI algorithms within the AIOps engine to fit their business parameters and their level of comfortability for automated network configurations based upon dynamic network or service conditions.”

NETSCOUT has a clear approach to help CSPs navigate this transformation. “We don’t just move data around in a garbage-in-garbage-out model; we curate it and feed it to the AIOPs engine. We deliver the right, real-time metrics but avoid a data bloat environment. We collect the data that the AIOps engine needs to run efficiently and make decisions in almost real-time. This cannot happen if the AIOps engine has to sift through terabytes of data and does not have telecom-level knowledge of what is relevant. This is why we came up with this approach: we want to provide a clean data stream based on what we heard about the need for visibility and to drive new initiatives from our customers.”

For more on AIOps for CSP’s visit:  
Next-Level AI for Telecom Networks | NETSCOUT