Juniper Networks is building out its artificial intelligence (AI) capabilities to make it easier for enterprises to manage the wired and wireless networks in their campus locations and while giving administrators more insights and automation tools.
The expansion of the vendor’s Mist AI portfolio and Marvis platform this week is part of its larger Experience First Networking initiative, aimed at shifting the focus of network management from how the network is running to the experience of both users and operators. AI is a key part of the strategy.
“The complexity is going up — more devices, especially as you see IoT [Internet of Things] coming into play, adding to mobility,” Christian Gilby, senior director of product marketing for Juniper’s AI-driven enterprise, told Enterprise Networking Planet. “I want to make it easier for IT teams to manage the network and understand what’s going on. It really spans our portfolio for the enterprise. … Ultimately, the outcomes we’re driving toward are really about assured experiences. It’s automation and ultimately evolving to a self-driving network. I don’t think everything will be self-driving and having the network just running itself, but there are certainly areas where we can leverage that to make it easier to operate the network.”
Buying Mist for AI
Juniper rapidly expanded its AI capabilities when it bought Mist in 2019 for $405 million, with the goal of infusing AI throughout its network portfolio. At the time of the acquisition, Mist was focused on leveraging AI for wireless LANS, but Juniper officials saw an opportunity to push AI throughout the IT stack and to advance the vendor’s management solutions. It also dovetailed with Juniper’s cloud-first ambitions, which have been amplified by the COVID-19 pandemic and the ensuing accelerated shift to cloud services as IT environments have become even more distributed.
As enterprises look at a post-pandemic IT environment, they’re seeing a new hybrid work trend and even considering reducing their central office sites and adding satellite offices, Gilby said. The key for this will be agility, which is what Juniper is aiming to drive with the updates to Mist AI and the Marvis AIOps platform.
The vendor is eyeing an AI-driven management of campus networks, which are under pressure from the fast-rising numbers of mobile users and IoT devices. EVPN-VXLAN technology can help with the fabric management by being an overlay sitting atop the physical network infrastructure, but it also is complex to both deploy and manage, according to Juniper. The company is aiming to enable EVPN-VXLAN management through its Mist Wired Assurance cloud service for wired, wireless and WAN.
Also read: Bringing Hyperautomation to ITOps
Easing EVPN-VXLAN Management
Juniper’s EVPN-VXLAN fabrics use the same technology — Juniper Mist Wired Assurance and the Marvis Virtual Network Assistant — that the vendor offers for managing wired access in the campus. The offerings give network administrators improved automation, insight, and troubleshooting capabilities. The aim is to bring a more modern management capability to an architecture that wasn’t designed for what’s happening in IT today.
“Mobility wasn’t really big when the architecture was designed; IoT wasn’t even a thing,” Gilby said, adding that the Mist AI capabilities are about the “need for a dynamic environment. You’ve got more devices connecting into the network. You’ve got to be able to secure it. When you look at a lot of these security breaches in the past few years, it’s been IoT devices because a lot of shortcuts are coming in because of the cost point that people are hitting. You’ve got to be able to segment the network and that’s where the complexity comes in. VLANs work to a degree, but they just don’t scale, so we’re starting to see a need for micro-segmentation. Also, how do you secure it? ACLs have been traditionally used, but again, it’s a very static way to think of the network.”
The Juniper EVPN-VXLAN fabric offers a common operational framework for all wired access deployments, configurations and Day 2 operations and a single AI-driven support model that includes self-driving operations and predictive actions. By moving EVPN-VXLAN into the Mist Fabric Manager, administrators can perform such tasks as choosing a topology, defining networks of interest, identifying physical connections and applying underlying policies.
For segmentation, Juniper opted for standards-based IPv6 rather than developing its own overlay and then taking advantage of the Juniper Mist Cloud to simplify the process of bringing a switch into the network, Gilby said. The company also is using its AI capabilities to help address service-level expectations (SLEs) by understanding the traffic going over the network through metrics and telemetry to detect and address issues and anomalies before they become problems.
“The area that a lot of vendors have not really focused on is operating the network and that was the key with Mist,” he said. “When we came out on the wireless side, it was really focused on, ‘You can get the network designed and deployed, but you’ve got to make it easier for the team to actually manage the network, get the insight to know what’s going on, because that is where they spend a lot of their time.’”
User Experience is Key
That’s a key differentiator for Juniper, according to Gilby.
“A lot of the vendors are talking about AI and algorithms,” he said. “What we really focus on is what AI can do for a customer. That is where all of us as vendors really need to move, teasing out what the customer benefit is, because it’s great to say you’re being forward thinking and using AI, but how does that translate into an outcome for the business? We’ve been there and will continue to do more, because that is what’s going to accelerate customers moving towards it.”
Marvis is Juniper’s AIOps network assistant and Marvis Actions is a dashboard used to monitor and manage the health of a Juniper Mist network, including wired, wireless and software-defined WAN (SD-WAN) networks. It leverages insights generated by the Mist AI engine, such as identifying the root cause of a problem, and provides actionable recommendations to IT managers.
Juniper has been adding capabilities to Marvis since it was introduced in 2018 to address common problems like misconfigured VLANS, non-compliant software and offline devices. The company has added nine new actions to address such issues as persistently failing wired and wireless clients, bad cables, coverage holes in access points (APs), bad WAN links, and insufficient RF capacity.
Coverage holes can be a challenge for enterprises, Gilby said. An access point might be placed in an area where there is interference with the signal. Marvis is designed to detect the user experience.
“We actually look at not just the AP-to-AP connection — all APs will listen to see if they have a peer and build a map — we also look at user experience,” he said. “Maybe the APs can hear each other, but there’s some metal ducting that’s blocking the signal for the clients. That’s where it’s important to also look at the user experience to detect these coverage holes.”
Read next: The Future of Network Management with AIOps