Networking

AI Self-Driving Networks Move From Vision to Reality

HPE Networking is using artificial intelligence, network telemetry and years of customer data to move networks beyond automated troubleshooting toward autonomous detection and remediation.

Bob Keaveney

Bob is the managing editor of BizTech magazine.

The promise of the artificial intelligence self-driving network is straightforward: Instead of waiting for IT staff to identify a network problem, diagnose its cause and implement a fix, the network can increasingly do those things itself. But getting there requires more than adding AI to a traditional network management platform. It requires years of telemetry, a deep understanding of network behavior and enough confidence to let AI take action.

That is the approach HPE Networking is taking, according to Rich Carter, a national sales director with the company. HPE Networking began offering cloud-native networking architecture designed around microservices roughly a decade ago. As AI capabilities evolved, it began using network telemetry and customer trouble ticket data to identify recurring problems and determine how they could be resolved.

“Now the next step was, what do you do with that information?” Carter says. “We’ve gotten some great information from AI that allows the system to look at very common problems that happen on a customer network and actually have the system autocorrect.”

DISCOVER: How HPE Aruba Networking can support your modern cloud experience.

From AI-Assisted to Self-Driving Networking

The distinction between AI-assisted networking, which has been around for a while, and an AI-powered self-driving network is important. Traditional AIOps can identify anomalies, discover problems and recommend actions. A self-driving network goes a step further by executing certain corrective actions without requiring an administrator to intervene.

HPE Networking’s Marvis AI engine currently supports a set of self-driving capabilities across wireless, switching and WAN environments. The company deliberately limits autonomous actions to scenarios where it has a high degree of confidence that the prescribed response will resolve the problem.

Carter compares the progression to autonomous vehicles. A self-driving car can operate safely because it has accumulated enormous amounts of data about roads, traffic and driving conditions. Similarly, HPE Networking has spent years collecting network telemetry and customer experience data.

“We have this 10-year head start of data that we’ve collected,” Carter says. “So now, when we do the self-driving actions, we know with a really high confidence level it’s going to solve the problems that the system is identifying.”

Still, he cautions that self-driving networking is not an all-or-nothing proposition. Customers can choose which actions they are comfortable allowing AI to perform while retaining control over others.

Click the banner below to learn the benefits of modernizing your network.

x-networkmodernization-animated-2024-clickhere-desktop

x-networkmodernization-animated-2024-clickhere-mobile

Autonomous Networks Start Small and Scale

One example illustrates how that works. A surveillance camera might be connected to a network switch, but its switch port can become stuck even though the port appears operational. The resulting outage might go unnoticed until someone needs to review footage.

With the appropriate self-driving capability enabled, the system can detect the stuck port and automatically cycle it, restoring connectivity to the camera without waiting for an administrator to discover the problem.

That kind of targeted remediation can be particularly valuable for organizations with lean IT teams. But Carter says that self-driving networking doesn’t need to eliminate network engineers; rather, this will reduce the amount of time they spend troubleshooting routine problems and give them better visibility into the problems that require human judgment.

Organizations new to autonomous networking can also start with a limited proof of concept rather than replacing an entire environment. HPE Networking may, for example, target a building with particularly poor Wi-Fi performance, install its equipment and measure the results in the production environment.

Carter says customers can then compare metrics such as trouble ticket volume, resolution times and network operating costs against their existing environment. The result is a gradual path toward autonomy: first gaining better visibility, then using AI to identify problems more quickly, and ultimately allowing the network to resolve selected issues on its own.

“Instead of the vendor that’s kind of, ‘Oh, me too, I’m going to bolt a large language model on top of my already existing solution,’” Carter says, organizations should look for networking platforms in which AI is fundamental to the architecture. “When we say AI-native, this is how the architecture started.”

For HPE, that self-driving vision now extends across both HPE Networking and Juniper Mist, following HPE’s acquisition of Juniper Networks last July. The goal, Carter says, is the same on both platforms: Use AI and accumulated network intelligence to move IT teams from reacting to network problems toward preventing and automatically resolving them.

Brought to you by: