Juniper AI-Native Networking Platform Solution Brief

These aren’t the kind of questions that can be addressed by bolting on new capabilities, AI or otherwise. They require long-term strategic foresight.

Juniper built the industry’s first AI-Native Networking Platform from the ground up to take full advantage of the promise of AI and make every connection count. Our AI-Native Networking Platform delivers the industry’s only true AIOps with unparalleled assurance in a common cloud, end-to-end across the entire network. Enterprises can rely on it to significantly streamline ongoing management challenges while assuring that every connection is reliable, measurable, and secure. Or they can build highly performant and adaptive network infrastructures that are optimized for the connectivity, data volume, and speed requirements of mission-critical AI workloads.

It all started with a strategic pivot to an experience-first approach that focuses on asking the right questions (Table 1) to deliver the best experiences for both network operators and end users. The ability to deliver the right experiences is built on three fundamental pillars.

• Right data
• Right real-time responses
• Right secure infrastructure

In traditional network infrastructures, green lights indicating connections are up only tell a small part of the actual story. In reality, some users may be unable to load emails. Others may be frustrated with choppy video conferences. And the list goes on and on.

Using AI as a solution for addressing granular network issues is a logical progression, but without the right approach it’s not a slam dunk. High quality data is essential. Juniper is the industry’s first technology provider to start with the right questions to arrive at the right data. These questions aim to uncover information about the user experience—along with top issues plaguing an organization’s network—as opposed to basic questions about connectivity status.

Juniper has over seven years of experience developing and relying on AI and ML algorithms that deliver high-quality insights with minimal false positives. The process of getting data right starts with our organizational structure and the teams responsible for building our solutions. At Juniper, our Customer Success teams collaborate closely with our data scientists on ways to enhance network experiences. With the right insights about customers most pressing ongoing challenges, Juniper data scientists can develop AI Primitives specifically designed to inform and drive the AI toolkits. To train our algorithms, we extract real-time telemetry data from routers, switches, access points, and firewalls.

It’s important to note that the Right Data concept extends beyond the network itself. Through continuous user experience learning, Juniper incorporates application data alongside network data, enabling the AI system to learn about the applications being used and predict potential impacts on a user's application experience based on adverse network conditions.

Ongoing struggles to respond to trouble tickets or to troubleshoot network issues are par for the course for networking IT teams. The AI-Native Networking frees up teams to focus on more strategic priorities by making issue detection and resolution a proactive ongoing exercise. Juniper AI models are trained based on context of anomaly detection, service levels, and event correlation, continuously improving themselves to provide IT teams with the right real-time response every time.

For example, Juniper’s Mist^™ AI has uncovered network issues and anomalies that have been affecting customers’ networks for months. Here are several examples illustrating how Juniper helps customers achieve the right real-time response using AI:

• Auto Radio Resources Management (RRM): Juniper’s Auto RRM leverages reinforcement learning to autonomously adjust access points’ channel and power in real time to optimize Wi-Fi coverage in a specific area or an entire building.
• Marvis^® Virtual Network Assistant: Marvis has revolutionized how network IT teams manage their networking infrastructure. Through a chat interface, IT operators can use natural language prompts and commands to troubleshoot issues and uncover underlying problems. Marvis is proactive in identifying real-time issues that affect user experiences and presents them in an easily digestible graphical user interface (GUI). IT operators can simply ask questions like, “Are there any unhappy users?” or “Troubleshoot Zoom for Kumar at noon today.”
• Marvis Minis: Also known as “Digital Experience Twins,” Minis simulate users and applications on the network. Even when users are not actively accessing the network (such as weekends or evenings), Minis continually monitor and identify configuration issues or other potential problems that could affect user experience.
• Service Level Expectations (SLE): At any given time, SLEs provide direct visibility into the true end-user experience across Wi-Fi, wired access, WAN, and applications. At the top level, SLEs describe exactly what the experience looks like. By double clicking, users can drill down into key details impacting the user experience. This includes correlations with other events or issues, as well as a description of the scope of impact, or “blast radius,” among other details—all in real time.

Marvis, the first and only AI-native virtual network assistant, simplifies operations for network IT teams by enabling real-time, proactive responses to network issues. Rather than manually troubleshooting issues, IT personnel can simply ask Marvis for help or look at the Marvis Actions framework for a summary of current issues and recommended actions. Marvis features include:

• Conversational interface: Marvis uses natural language processing (NLP) with natural language understanding and knowledge graphs to understand user intent and goals. It contextualizes inquiries, returns specific results—and can even take actions based on user feedback.
• Marvis Actions: By leveraging the Mist AI engine to identify the root cause of issues across WLAN, LAN, WAN, security, and data center domains, Actions can automatically fix (self-driving) or recommends actions (driver-assist) with high efficacy.
• Marvis Minis: The only AI-Native Networking Digital Experience Twin, Minis proactively simulate user connections digitally to instantly validate network configurations and find/detect problems without users being present. Data from Minis is continuously fed back into the Mist AI engine, providing an additional source of insight for the best AIOps responses.
• Anomaly detection: Marvis adds anomaly detection to the SLE framework to enable administrators to proactively identify service-impacting events and assure rapid determination and resolution of issue root causes.
• Data correlation: Marvis correlates information across a large knowledge base to determine the scope and magnitude of a problem.
• Accurate root cause analysis: Our data science toolbox uses Bayesian inference to identify and pinpoint the highest probable causes for network issues.

From devices to operating systems to hardware to software, Juniper has the industry’s most scalable infrastructure, underpinning and supporting our networking platform. Our true cloud-native, API-connected architecture is built to process massive amounts of data to enable zero trust and ensure the right responses in real time.

Juniper Mist Cloud provides the backbone for the Juniper Mist AI engine, centralizing data and AI compute with the scale and elasticity to train AI and execute AI inference. Juniper Mist Cloud delivers updates, feature releases, bug fixes, threat patches, device support, and much more in real time with no interruption to network operations. Customers enjoy simpler operations with no need for cumbersome on-premises hardware and software additions.

The Right Secure Infrastructure also encompasses Juniper’s ability to provide the specific networking infrastructure required for AI clusters. We have a proven track record in delivering robust, high-performance, standards-based infrastructure, and we’ve committed to ensuring our Ethernet-based networks meet the demanding needs of enterprises building data centers for AI training, storage, and inference.

Beginning with our early leadership in wireless LAN, Juniper continues to extend AIOps to all parts of the network.

Transform campus and branch networking with the industry’s only AI-driven, cloud-native campus and branch solution proven to deliver predictable, reliable, and measurable end-user experiences with a low TCO. The AI-native full-stack solution provides AIOps for wireless, wired access to core, SD-WAN, indoor location services, and network access control (NAC) for a transformative, AI-driven full-stack experience from client-to-cloud (see Figure 4). Specific use cases include:

• AI-driven full stack
• SD-WAN
• SD-branch
• Enterprise edge
• SASE
• Asset tracking
• Mobile engagement
• Wired and wireless access
• Wi-Fi access
• Wired access
• Campus fabric
• Private 5G

Embrace the hybrid cloud and AI revolution with an assured, secure modern data center that empowers IT teams to scale innovation more easily and reliably. Intent-based data center automation and continuous validation from Day 0 through Day 2+ with Juniper Apstra^® software and Marvis simplifies operations. Use cases include:

• Data center automation
• Data center fabric
• Data center edge and WAN gateway
• Zero trust data center
• Data center interconnect

Effortlessly scale automated WAN for optimal experiences, making the network itself imperceptible. Juniper's WAN Edge Routers deliver unparalleled versatility combined with minimal carbon footprint. Blending proven performance with simplified AIOps and automation, Juniper WAN Edge Routers ultimately empower enterprises with the sustainable and future-proof network required for digital transformations.

Use cases include:

• Private WAN Backbone
• Internet Peering
• Cloud Connect
• Data Center Edge
• Data Center Interconnect

Juniper Mist Cloud uses a modern microservices-based design to bring unparalleled agility, scale, resiliency. This makes it easy to add or remove new features, implement new enhancements, and deliver bug fixes almost weekly without network disruption. Services scale up or down elastically when they’re needed, eliminating the cost and complexity of monolithic hardware. Plus, Juniper Mist Cloud is inherently resilient as the failure of one service does not impact others.

Juniper’s AI Data Center solution is the quickest and easiest way to deliver high performing and scalable networks for AI training and inference. With unique intent-based operations, an AI-native virtual network assistant, and Juniper validated designs, Juniper takes much of the complexity out of AI Data Center networking design, deployment, and troubleshooting, allowing enterprises to do more with fewer IT resources.

Apstra enables intent-based networking across multivendor environments, with templated blueprints, continuous validation, and proactive troubleshooting. It also integrates with the Marvis Virtual Network Assistant, which provides AI-driven analytics and insights for data center operations.

Juniper Networks^® QFX Series Switches and PTX Series Routers deliver high-bandwidth, lossless, low-latency, and scalable connectivity over Ethernet, which is the preferred technology for AI data center networking. Ethernet leverages the vast ecosystem of vendors and standards, driving down costs and accelerating innovation.

PTX Series Routers provide high-density 800GbE platforms for spine and super spine architectures, supporting thousands of GPUs at the highest speeds. Juniper also offers validated designs for AI clusters of different sizes and configurations, as well as comprehensive security solutions and integrations with third-party vendors.

To extend assurance beyond the network to applications and services, Juniper uses an open API ecosystem. It all started with joining Zoom data and network data and using our continuous user experience learning process to get to the bottom of what causes poor Zoom experiences (see Figure 6). We used the learnings to help address the issues before users even connect, effectively extending end-to-end assurance beyond the network to applications and services.

AI-Native and sustainability initiatives go hand in hand. By enabling fast and remote troubleshooting, AIOps helps organizations reduce carbon-heavy site visits by up to 85%. Moreover, Juniper networking infrastructure is designed with an eye on the future; we prioritize power efficiency and modular designs that simplify repairs and reduce e-waste while extending product lifetime. And we ship our products in sustainable-minded packaging designed to reduce waste and shipping emissions.

Juniper laid the foundation for the AI-Native Networking Platform years ago when we had the foresight to build products in a way that allows us to extract rich network data. By using our data to answer questions about how to consistently deliver better operator and end-user experiences, and building the capabilities that allow us to do that into our DNA, we’ve set a new benchmark for the industry. And we’re not planning to rest on our laurels. Enterprises looking to deliver exceptional end-to-end user experiences while simplifying networking challenges and assuring security need look no further than Juniper to make every connection count today and tomorrow.

Learn more about the Juniper AI-Native Networking Platform.

3510815 - 002 - EN JUNE 2024