Masergy announced a series of improvements to its artificial intelligence operations (AIOps) platform today in an effort to improve software-as-a-service (SaaS) and network performance.

The managed service provider announced its AIOps platform in November 2019 as a “virtual engineer” tasked with sussing out network anomalies and serving up predictive analytics. Since then the service has become a value-added feature for Masergy’s SD-WAN and secure access service edge (SASE) customers.

“The goal all along has been not necessarily to provide prescriptive recommendations about the network, but to be able to provide prescriptive recommendations and even implementations about applications,” Ray Watson, VP of innovation at Masergy, said. “At the end of the day, nobody cares what your packet loss is. They care how your applications perform.”

To achieve this, Masergy started by training its AIOps platform with more than 20 years worth of network, security, and application data.

“We put 20 years of data in there. We only put the good stuff in. So we took out all the trouble tickets that were either not resolved with customer satisfaction scores that looked right, or that we weren’t highly confident that the resolution was actually the resolution,” he said. “One of the great things in the world about being in telecom is sometimes someone reboots a router, and it just works and we have no idea why.”

Watson adds that the model is continually being trained to understand what’s normal for a customer's network environment. “If you have one location in Wimbledon, United Kingdom, that has duplex mismatch errors every other Wednesday, it knows that so you can immediately make predictions based on that,” he said.

This is then broken out by the customer’s top 20 applications and graded based on known acceptable metrics for things like packet loss, jitter, or latency, which can be tweaked on an individual basis. Customers are then fed recommendations based on the network behavior.

“Instead of just telling you things like [dynamic host configuration protocol] mismatches, or [quality of service] misconfigurations, we're also going to be able to give you bandwidth predictions, bandwidth recommendations,” Watson said.

If an application is experiencing high latency, for example, Masergy’s AIOps platform might recommend rerouting traffic to a different geographic point of presence and using forward-looking routing will predict latency savings. The success or failure of these recommendations are then used to further train the AI model to provide more granular insights.

Watson notes that Masergy has yet to fully close the loop on the AIOps platform. He said the company is still reviewing recommendations for accuracy in order to prevent downtime from a bad interpretation of data.

“Unplanned downtime is still largely due to manual processes and human error. AIOps eliminates these challenges, revolutionizing IT operations,” Zeus Kerravala, founder and principal analyst at ZK Research, said in a statement. “The value of Masergy’s AIOps stems from its ability to evaluate bandwidth usage patterns, identify anomalies, and predict outages all within a fully managed SD-WAN or SASE service.”

Masergy’s AIOps platform is available now as part of the service provider’s SD-WAN and SASE offerings.