Big Data From Enterprise AI Will Unleash The New Era Of AI-Driven WAN And AIOps

Big Data From Enterprise AI Will Unleash The New Era Of AI-Driven WAN And AIOps

Big Data From Enterprise AI Will Unleash The New Era Of AI-Driven WAN And AIOps

This year, with many enterprise AI initiatives and distributed networking capabilities promising more resiliency, the fourth generation of network infrastructure integrates SD-WAN with AIOps. With Juniper Network’s acquisition of 128 Technology, companies see the future beyond SD-WAN to AI-Driven WANs. As AI-enabled technology generates mountains of Big Data, lowering the cost of infrastructure maintenance and network overhead is a must.

I had the pleasure of interviewing Bob Friday, CTO of MIST Systems, the AI company behind Juniper Network’s Intelligent Network Solutions, such as the Marvis Virtual Network Assistant, WAN Assurance Software, and more. MIST AI is the brain behind Juniper’s AI-Driven WANs and AIOps.

Conversational AI Is Powering AIOps

Conversational AI is not only transforming customer service. It is also changing IT Support. Network Support historically required a handful of network engineers monitoring and supporting a vast amount of network infrastructure. With Enterprise AI initiatives, data and devices multiply exponentially, and WAN innovation has to keep up.

Bob Friday says, “A few years back, when I was in Cisco, I saw a paradigm shift from managing devices to customers wanting end-to-end visibility. It wasn’t good enough to know that the network was working. But, they really wanted their customer or device to know that the connectivity was good. We decided to build something that can answer questions and manage the networks on par with domain experts. Now, our AI technology is getting to a point where it can answer very difficult questions like, why does a person have a bad internet connection?”

Recently, the industry’s move to SD-WAN was also to lower cost and increase transparency. Support engineers needed a lot more efficient ways to diagnose and troubleshoot problems on the network. AIOps using an AI assistant to monitor and analyze network problems allows for a more proactive support model.

Bob Friday says, “The other paradigm shift is going from CLI to dashboards. I think right now, networks are getting so complex. These dashboards are even too much data for the average person to deal with. There has to be a better way for Network Support personnel to interact with the network. You want the AI assistant to answer these questions and become a member of the support team.” 

Bob Friday says, “Conversational interface for me was a kind of guidance. It gets rid of the dashboard and provides an easier way to get data.”

Trust Is The Key To Responsible Implementation of Conversational AI Interface

From the beginning, it was apparent to Bob that trust was a big issue in incorporating the conversational AI interface into a network support team’s day to day work. AI is good at spotting patterns and establishing strong correlations. Troubleshooting standard network problems immediately points to strong cause and effect relationships. If a server is down, then chances are you are going to have connectivity problems.

Bob Friday says, “If you hired a new person, you’d ask, how do I trust this person? Our AI assistant has to demonstrate how it got the answer right. You are having connectivity problems because your DHCP server is down. So, you have to present some evidence to gain trust. Why would someone trust the AI assistant any more than a new hire?”

Bob and his team started developing the AI’s capability to figure out time-consuming problems for support personnel.

Bob Friday says, “We’ve gotten to the point now where we can probably 97% accurately detect connectivity on ethernet cables. That’s something that would take a person a long time to figure out. That’s the type of problem for AI to solve.”

The key for AI is the feedback loop that allows it to improve from learning to solve simple problems and solve more challenging problems. Unlike a human engineer, AIOps can collate vast amounts of information from different network nodes quickly to solve a problem.

Bob Friday says, “In the cases where Marvis didn’t quite get the right answer, once the human sees the data, they can say that you got close. It’s a quicker way for the administrator to access the data rather than logging into many machines and going through all of these dashboards. Marvis can get the administrator closer to the solution. ‘Hey, it’s these three things. I found three possibilities.’ Then, the administrator can figure out the rest.”

Trust also comes from the type of data that AI is ingesting. In the network support area, data related to networks are structured data that fits a logical framework. AI can make good decisions based on a logical path of inference.

Proactive Support Reinforces Accountability

AIOps adds a different layer of transparency to the support model. Historically, IT support had a hard time dealing with vendor infrastructure and software. Now, AIOPs can allow a more proactive diagnosis of potential issues. This proactive environment enforces a new type of accountability.

Bob Friday says, “AIOps is almost vendor-agnostic. Before, you typically had to log into a router or switch to figure out what’s going on. But, now you have AI and all of that data, customers don’t have to wait to tell us there’s a problem. Customers can say that we know there’s bad hardware or software in your network. So, customers don’t have to fight with their vendor to get them to admit there’s a problem. You can just send a bunch of evidence. AI turned that full support model upside down.”

To figure out the solution to a problem, AI can comb through more data from various sources. For instance, to answer a question about the availability of a data center, there are many factors. AI can comb through data from many sources quickly and more effectively.

Bob Friday says, “If I’m trying to answer a question about the data center, eventually we want to be pulling data from the weather forecast. What’s the probability of these servers overheating? Well, so it’s 120 degrees outside, you don’t have enough air conditioning, you may want to start predicting that. AI starts with what question you’re trying to answer. You tend to want to pull data from as many sources as you can to answer that question.” 

Edge AI and increased IoT sensor data on a complicated distributed architecture make AIOps almost essential going into the future. Endpoint security is one of the biggest concerns as more devices connect to the network.

Bob Friday says, “I think what you’re starting to see is, as they spot  more sensors on the network, that don’t have endpoint security on them, they’re looking to AI to help keep an eye on what are all these headless devices doing on the network?”

Bob Friday says, “Security has been an early adopter of AI. If you look at the user entity, behavior analytics, that segment of the market, they’ve been an early adopter of trying to use ML machine learning and AI to spot anomalous behavior and traffic flows. You can try to create risk scores. On the connectivity side, we’re creating connectivity scores on, how good is your connectivity experience?”

The Future of Increased Connectivity

AIOps and AI-driven WAN will help to support increased connectivity of mobile devices all over the world. With the advent of 5G, streaming will become ubiquitous. New types of emerging technology that require reliable, low latency connections will emerge.

This article originally appeared on forbes.com To read the full article and see the images, click here.

Nastel Technologies helps companies achieve flawless delivery of digital services powered by middleware. Nastel delivers Middleware Management, Monitoring, Tracking and Analytics to detect anomalies, accelerate decisions, and enable customers to constantly innovate. To answer business-centric questions and provide actionable guidance for decision-makers, Nastel’s Navigator X fuses:

  • Advanced predictive anomaly detection, Bayesian Classification and other machine learning algorithms
  • Raw information handling and analytics speed
  • End-to-end business transaction tracking that spans technologies, tiers, and organizations
  • Intuitive, easy-to-use data visualizations and dashboards