Americas

  • United States

Mist’s AI-driven virtual assistant speeds up network troubleshooting

News Analysis
Jun 27, 20195 mins
Enterprise ApplicationsJuniper NetworksNetwork Management Software

Digital assistants using artificial intelligence will have a bigger impact as an IT support tool, as Mist Systems' virtual network assistant Marvis demonstrates.

ethical ai artificial intelligence algorithms
Credit: Iaremenko / Getty Images

The use of digital or virtual assistants and chatbots has picked up momentum with the rise of artificial intelligence (AI). These automated assistants have been around for years, but they haven’t been all that useful, as they required a significant amount of programming to look for certain keywords and then the responses were based on logical guesses. 

The infusion of AI, however, has made these systems much smarter and now natural language queries can be made and responses are in plain English (or other languages).

Many businesses have adopted digital assistants and chatbots to improve customer service. For example, Atom Tickets uses conversational AI to enable people to book movie tickets and even dinner with just a short sentence instead of having to go through the rigmarole of going back and forth with discrete commands.

AI coming to network operations

Despite the advancements with digital assistants, they have yet to impact the way IT departments manage the network. More and more vendors use AI to find security threats, configuration anomalies, and as a predictive tool, but it hasn’t fundamentally transformed the way companies troubleshoot and manage their networks.

Juniper hopes to change that, and this week the company announced several updates to its recently acquired Mist business unit. The announcement included new W-iFi 6 access points (APs) with integrated AI intelligence. Given Mist’s differentiator has been AI since its inception, this wasn’t a huge surprise. Juniper also announced a product called Mist Edge that enables some of the Mist services to be run on-premises in a hybrid configuration similar to the Google Anthos or how Amazon Web Services (AWS) Outposts would operate.

The most interesting part of Mist’s new capabilities are the updates to its digital network assistant, Marvis, to make it a more effective troubleshooting and support tool. The name Marvis is a play on words — a combination of Mist and Jarvis from Iron Man fame. In the Iron Man movies, Tony Stark is able to ask Jarvis questions, and the natural language processing enables Jarvis to respond.

Similarly with Marvis, if a user called the IT helpdesk and said an application such as Zoom isn’t performing well, the engineer could ask Marvis what’s wrong with Zoom and Marvis would use its AI engine to determine the problem.

Mist accomplishes this with a strong AI toolkit, but it has also added a capability called dynamic packet capture (dPCAP) that automates the collection of information when the troubleshooting process starts. Packet capture can be extremely useful, but it typically requires an engineer to deploy specialized equipment locally. The automation capabilities Mist brings enables packet capture to be done anywhere and anytime.

AI-based troubleshooting can address one of the biggest pain points with network operations

Using AI for troubleshooting and IT support can pay huge dividends. Most network engineers are smart and savvy and can fix a problem quickly once it has been identified. The problem is, particularly with Wi-Fi, finding the source of a problem can be very time-consuming. My research has found that 90% of the mean time to fix a problem is in the identification phase. Using Marvis to shorten that time addresses the main source of time consumption for many network professionals, as it streamlines the process of finding the root cause.

The reason why Wi-Fi is so hard to troubleshoot is that a number of things have to work correctly for the experience to be solid. Wi-Fi association, authentication, congestion, DHCP, DNS, and other factors play a role. If the problem is congestion or association, the issue may fix itself when clients move out of the area. Often, a user will call with a complaint, and the issue will resolve itself before the source of the problem is found — meaning the problem wasn’t fixed; it just temporarily went away. This is what users find so maddening about Wi-Fi – it works one minute, then stops, then starts working again. And that’s why some network professionals spend a day or two a week doing nothing but Wi-Fi troubleshooting.

Marvis is designed to eliminate this time suck through a conversational platform. Without the conversational element, the engineer could use Marvis to find the problem, but it could take a series of questions, like, “Marvis, what’s the performance of Zoom?” “What clients are connected to the AP?” “What’s the quality of video?” “What’s the quality of voice?” and so on until the problem is isolated. Juxtapose that with the conversational model where a simple question can be asked, and it’s easy to see the value that the enhanced, conversational model can bring.

AI delivers proactive network support and remediation

In many cases, the AI engine will actually notice there’s an issue before a user does. In this case, Marvis would proactively notify the IT staff there’s a problem and provide guidance on how to fix it. If the issue isn’t resolved automatically, the process of opening a trouble ticket can be automated. The goal of every organization should be that the operations team notices all problems before the workers do, but in actuality, 75% of trouble tickets are opened by the user and not the IT department. AI can help reduce that percentage down to zero.

The network industry is just starting scratch the surface of what’s possible with AI, and this will be an area for innovation for years to come. IT professionals should consider the AI as a smart tool to help operate day-to-day tasks more efficiently, enabling people to focus on more strategic initiatives. Conversational AI will prove to be a game changer in network operations, as it will allow first-level engineers to raise the questions and hand it to Marvis to find the problem and hopefully fix it without having to tap into more expensive and busier engineering resources.