The Marvis Virtual Network Assistant (VNA) is the digital network expert that supports your own team of network experts—network administrators, site reliability engineers (SREs), or anyone based in network operations centers (NOCs). Marvis is the first network assistant in the industry to bring conversational AI to networking, transforming the way IT teams interact and engage with enterprise networks.
In the era of AIOps, Marvis delivers streamlined operations, simplified troubleshooting, and remarkable user experiences. Since 2016, the Mist AI™ engine has applied various data science tools to continue to learn and improve, expanding its knowledge base as it becomes a fundamental component of The Self-Driving Network™.
A Conversational Assistant
Conversational assistants have come a long way to become a part of our everyday lives, evolving the way humans interact with computers—the most well-known ones being Siri, Cortana, and Alexa. They have formed an area of strategic investment across several industries, including banking, retail, and healthcare, as organizations look to streamline operations and deliver personalized experiences to users. And now, Juniper Mist™ is the first to bring a conversational assistant to enterprise networking, driven by Mist AI.
Marvis enriches its natural language processing (NLP) capabilities with natural language understanding (NLU) to offer a Conversational Assistant that understands user intent and delivers improved value and quality of returned results. The power of the Marvis conversational assistant is that it can contextualize requests to accelerate troubleshooting workflows, answer product or feature specific questions, provide information about the network, and help find any type of network device. Specifically, Marvis can:
- Get real-time answers about the network in a few clicks
- Deduce user intent from general statements and inquiries using advanced NLP with NLU and NLG
- Improve specific user experiences by learning from user feedback
- Ask generic questions beyond troubleshooting, like “How to setup RRM?” and “Does AP have capacity?”
Marvis significantly helps with troubleshooting and support so IT teams can get to quicker mean time to resolution or innocence. It presents a comprehensive network view with user, client, and device insights, eliminating the need to pull up multiple dashboards or memorize CLI commands. Marvis is an extension of the IT team—it combs through data and logs to determine root causes and provides answers in real time with high efficacy.
Marvis is the one-stop shop for IT teams to understand the state of the network. They no longer have to memorize CLI commands or know which dashboards contain the relevant information as it is accessible from any page in the Mist dashboard. With answers at their fingertips, it significantly transforms the way IT teams experience and interact with the network.
Client-to-Cloud View Via Actions: Journey to a Self-Driving Network
Marvis Actions is about driving operational simplicity and transforming IT from reactive troubleshooting to proactive remediation. It offers a “morning cup of coffee” view, which delivers visibility into high-impact network issues at an organizational level so that administrators know exactly what they need to prioritize and focus on for the day. As sites get added, Marvis Actions scales with ease as no additional setup is required from the user.
Marvis proactively identifies the root cause of issues across IT domains (WLAN, LAN, WAN, security, and applications), providing insights into every high-impact issue seen when a client is accessing an application either in the data center or cloud (client to cloud). It can either automatically resolve issues (self-driving mode) and in the future with permissions from the user, or it recommends actions that require user intervention (driver assist mode). Once completed, Marvis closes the feedback loop by validating that the actions are correct in the Mist AI engine, thus helping Marvis learn while gaining your IT team’s trust.
For Juniper Mist™ Wired Assurance, some Marvis Actions are listed here. Marvis isolates wired clients failing to connect, and it identifies bad network cables connected to ports, mismatched port settings, L2 loops, DHCP failure scopes, wired clients failing to authenticate, and continuous port flaps with an option to also disable the port that the user can ask Marvis to automate in the future.
It also automatically adds missing VLAN tags, fixes incorrect port mode configuration, and isolates persistently failing wireless clients. It also identifies missing VLANs from third-party switches. For Juniper Mist™ Wi-Fi Assurance, Marvis Actions helps isolate APs that missed a firmware upgrade in any given site, provides root causes and failure scopes when an AP goes offline (site or switch down scenario), isolates persistently failing wireless clients, detects coverage holes and APs or locations with insufficient capacity, identifies APs connected to a bad cable, and detects missing VLANs. Another unique capability is providing insights into issues with upstream services and devices such as radius, DHCP, DNS servers, and ARP gateways.
Root cause analysis not only includes server failures, but also includes complete failures seen on an AP, switch, WLAN or across a site or organization.
Additionally, Marvis realizes our AI-driven support with proactive return material authorization (RMA) for unhealthy Juniper access points. It eliminates the manual troubleshooting checkpoints initiated by support teams, delivering significant savings in time and effort, while boosting the overall experience of users, devices, and clients.
And with an API-driven interface, Marvis can trigger these actions via webhooks and email alerts for customers to auto open support tickets or work orders in their systems.
To learn more about some actions, visit: https://www.mist.com/documentation/category/marvis-actions/.
The Marvis client, currently available for Android and Windows devices, is a software agent that sits on the end-user device, collecting and displaying detailed client-device properties, including insight into client roaming behaviors.
Additionally, the Marvis Client recognizes device connection types—cellular or Wi-Fi—along with the corresponding signal strength. This additional level of granularity lets admins better understand the Wi-Fi experience directly from the client’s perspective.
The Marvis client further enriches the client device data. It goes beyond basic fingerprinting to provide an additional layer of detail, such as device type, manufacturer, and different versions of operating systems. The more details the client APK can draw out, the better the Mist AI engine gets at advanced device classification. Marvis continually learns and becomes more accurate in its ability to distinguish between device-specific issues or broad device issues, such as specifically identifying that OS version 8.1.0 is affecting certain clients.
Marvis incorporates anomaly detection within SLEs so administrators are notified of service-impacting events and can quickly identify and resolve the root cause of issues. Anomaly detection automatically applies machine learning to establish service baselines and triggers notifications when there is a deviation from known standards. The feature uses our third generation of algorithms in long-short-term memory (LSTM) and recurrent neural networks (RNN) to boost efficacy over 95%, minimizing false positives.
Juniper has expanded Marvis's conversational interface (CI) capabilities to deliver even more human-like conversational capabilities, particularly regarding documentation and support issues. More specifically, Marvis can now leverage a ChatGPT interface when looking for historical information from Juniper’s public-facing knowledge base information.
With new Zoom integration capabilities, Marvis is able to collect key user experience application data for Zoom clients and correlate it with key wired, wireless, and WAN data using advanced AI/ML techniques to identify the root cause of video conferencing problems quickly.
Zebra Wireless Integration
The Marvis client along with the Zebra wireless insights provides visibility into networking, connectivity, and application issues (voice) inline with when the device is in use. This provides the device Wi-Fi experience (as seen by the device) without running any additional synthetic tests.
Client Service Level Expectations (SLEs)
Marvis complements the client SLE framework. It applies machine learning to track and monitor client and device experiences with continuous behavioral analytics and network traffic analysis. The ability to understand these trends gives IT deeper insights for troubleshooting and planning.
Multivendor Switch Insights
IT teams can use Marvis to gather health statistics of both Juniper and third-party switches connected to Juniper Access Points, including:
- How many access points are connected to a switch
- PoE compliance status, which helps manage and balance the power draw of connected devices
- Identification of VLANs that are misconfigured on switch ports where APs are connected but clients are getting blocked
- Version compliance for switches running dissimilar hardware
- Switch uptime
About Juniper Networks
At Juniper Networks, we are dedicated to dramatically simplifying network operations and driving superior experiences for end users. Our solutions deliver industry-leading insight, automation, security and AI to drive real business results. We believe that powering connections will bring us closer together while empowering us all to solve the world’s greatest challenges of well-being, sustainability and equality.
1000698 - 003 - EN MAY 2023