How ML and AI Help Optimize Network Management
The term machine learning (ML) was first coined in 1959 by Arthur Samuel, an artificial intelligence (AI) pioneer who developed “computer checkers,” one of the first gaming programs. But ML/AI didn’t enter into broad conversation until the last several years, as ML shaped the technology-based capabilities we’ve now grown accustomed to – like autocorrect and search page rankings. At Hughes, our engineers leverage ML and AI across our networks to improve the customer experience.
“Machine learning is a technique to forecast and predict events based on historical data that has been collected over time,” explained Archana Gharpuray, vice president of software engineering at Hughes. “This data is then used to train a mathematical model which can predict future events with a reasonable degree of accuracy. AI takes it one step further. AI determines what automatic action can be taken when a machine learning agent detects an anomaly.”
At Hughes, ML/AI innovations are in use across our increasingly complex customer networks, helping to optimize the explosion of cloud-based services, diverse hybrid environments, mobility services, and the rise in Software Defined Wide Area Network (SD-WAN) technologies.
The burgeoning amount of available data – from the cloud, billions of connected devices, and the Internet of Things – can now be digested by ML models and applied to network management tasks. The ML agent can prompt network engineers and customer service agents to resolve an issue before a problem occurs, or in some cases, AI can determine an automatic action to preempt the event.
With ML/AI, network monitoring is no longer an “on demand” activity, where the customer calls about a problem and someone troubleshoots the issue. Instead, at Hughes, ML in the cloud monitors customer networks in real-time, all the time. In that way, for example, ML can predict network congestion and assist with traffic classification. For enterprise network management, this is critical for such tasks as capacity planning, security and intrusion detection, Quality of Service, performance monitoring, application prioritization, and efficient resource management – all of which help deliver a better customer experience.
ML can also be applied outside of day-to-day network operations and free up the network engineer to do higher value tasks.
“We’re using AI and ML to see what transports might be available that are most aligned to the business needs at a retail location—whether there’s cable, fiber or DSL,” explained Bill Rumancik, senior director, enterprise solutions and marketing at Hughes. “We’re also using ML to predict if a satellite terminal is going into a degraded state, even before the terminal can pick up on it and report it. The algorithms send an alert to a network engineer to solve the problem, which leads to higher customer satisfaction, because they didn’t experience an incident or outage.”
Hughes recently announced that it is making a new AI for IT operations (AIOps) feature commercially available for enterprise wide area networks. This AIOps feature has already been in use across 32,000 managed sites, as part of our HughesON™ Managed Network Services. We use AIOps to predict and “self-heal” network anomalies before they can cause service disruptions, with roughly 70% of cases being corrected automatically.
In other cases, when the ML agent detects an anomaly, like a problem site, it sends an alert to a Hughes team member using a variety of formats. There can be alerts in the dashboard, or via email, chat bot, and even a voice personal assistant (like Alexa and Siri). The Hughes team member can then interact with the ML agent to learn more details, with the ML agent pulling additional data from the cloud to provide more information and answer questions. Essentially, the process becomes an interactive triage session.
For consumer networks, like Hughesnet®, ML/AI network monitoring identifies certain types of traffic – such as video – and applies a data-saving feature to maximize the customer’s data plan each month. We also use AI to ensure Hughesnet installations are completed correctly. Our installers send photos of completed installations, and the system automatically examines them for any issues that need to be addressed.
Actively incorporating the latest innovations in ML/AI, Hughes streamlines network management, improves efficiencies, and enhances the customer experience.