Partner POV | What Does It Really Mean to Be AI-Native?
Article written by Rami Rahim Chief Executive Officer of Juniper Networks
Walk the halls at CES. Listen to the conversations at the World Economic Forum in Davos. Just open your own social media feeds.
AI is taking over.
When OpenAI's ChatGPT burst onto the scene in November 2022, its human-like conversational interface and ability to answer even the most nuanced questions finally made AI real for many people. And as the platform hit a record 1 million users in the first 5 days, a surge of companies began rolling out their own AI-driven or AI-powered tools and products to ride the wave.
I believe AI truly has the power to transform nearly every aspect of our lives. And when the AI-hype dust settles, we will quickly see that a new era of computing, working and living has taken shape.
Luckily, at Juniper, we've known real AI for years.
Our Mist AI solution has been using a combination of artificial intelligence, machine learning and data science techniques to drive experience-first networking for customers all over the world since 2015. ServiceNow, for example, saw a 90% reduction in network trouble tickets using Mist. Gap, Inc., reduced truck-rolls to their stores by 85% with Mist resolving IT issues on its own.
Today, we're taking it to the next level.
With the industry's first AI-Native Networking Platform, we're extending AI throughout our whole portfolio to assure experiences and simplify operations end-to-end across every network domain.
What Is the Juniper AI-Native Networking Platform?
Leveraging seven years of AI learning, our closed loop AI-Native Networking Platform brings end-to-end assurance across the wired access, wireless access, WAN, security and data center domains with common cloud-hosted AI Operations (AIOps) and a Virtual Network Assistant (VNA), called Marvis.
AI-Native networking isn't just a fancy phrase, framework or "market-tecture." This is a platform that drives experience-first networking better than any other solution on the market because real, proven AI was built into the original product architecture.
What we have now are networks built for AI. Optimized for AI. From the very beginning.
How do we go about building this platform?
We start by asking the right experience-first questions. I don't mean questions like, "Is the network up?"
Rather, questions like, "How do we ensure every user, in every location, is getting a consistent experience?" or "Can the network adapt to fix issues before users even know they exist?" or "Does the network meet the needs of all of its stakeholders?"
Because for too long our industry has focused on metrics like the performance of devices, the status of network infrastructure, the speeds and throughput of equipment. These are important pieces of the puzzle, but the puzzle has gotten much bigger and more complex in the last several years. And the winners of the next decade of networking will be those who understand that the focus must revolve around delivering flawless experiences.
By starting with these experience-first questions, we then go about answering them by having the right technical ingredients, including the industry's only true AIOps that is capable of running across the entire end-to-end network.
But not all AIOps is created equal. At Juniper, we believe world-class AIOps comes down to three key aspects.
First, it's about having the right data. Because you can't fix what you don't know needs fixing.
Fortunately, we had the unique foresight years ago to build products in a way that makes it possible to extract rich network data from them. So now, we're collecting mountains of the right data from switches, access points, routers and firewalls that allow network operators to understand what end users are experiencing.
Of course, it's not enough to just collect the right data. It's also about what you do with it. The right real-time response must be actionable and go beyond flashing alarms. By understanding the satisfaction of every user on the network, at all times, we solve problems for customers before they ever realize there's an issue.
See, bad AIOps gives you flashing lights and alarms – often false ones that IT administrators eventually just ignore. Good AIOps – Juniper's AIOps – is about proactively fixing issues before they start affecting experience.
Lastly, none of this is possible without the right secure infrastructure.
Juniper has the industry's most scalable and secure infrastructure for the AI era—from devices to operating systems to cloud hardware and software. It's infrastructure that's built to process massive amounts of data and, importantly, achieve game-changing outcomes with it.
Real AI depends on these three ingredients. It must have these three ingredients: the right data. The right response. The right infrastructure. The result is unparalleled agility, automation and assurance to reduce complexity, increase productivity and ensure reliable performance at scale.
And make no mistake – this isn't a roadmap. Juniper's AI-Native Networking Platform has been seven years in the making, and it's here now with real products in this real platform.
We even have several new innovations that we're introducing today under its umbrella.
AI for Networking
First, we're expanding and extending Marvis, our virtual network assistant that's at the heart of our AIOps solution, in two important ways.
The new Marvis Minis are essentially our digital experience twins, which find and solve problems before users know they exist. Typically, an AI system needs users present on the network and generating data to troubleshoot and resolve any given issue. With Marvis Minis, that need goes away.
We're also bringing Marvis to the data center, which is going to yield unprecedented levels of automation in this domain. Juniper® Apstra can already automatically configure a network based on outcomes you specify, and then continuously validate that operations match the intent. But if something ever goes wrong in the data center that negatively impacts end-user experience, that's where Marvis now comes in. And in the same way that Marvis reduces network trouble tickets in the campus and branch, we expect it to meaningfully drive down issues and reduce resolution time in the data center.
And speaking of the data center, the other important aspect of AI lies in how we architect data center networks to support the massive new requirements to train AI models and process AI workloads.
We call this…
Networking for AI
As I've written before, modern AI/ML clusters comprise hundreds or sometimes thousands of GPUs in a data center that provide the massive, parallel computational power that is required to train today's AI models. Distributing workloads across GPUs and then syncing them to train the AI model requires a new type of network that can accelerate job completion time (JCT) and reduce the time that the system is waiting for that last GPU to finish its calculations (tail latency).
With new 800G-capable PTX routers and QFX switches to complement Apstra and Marvis in the data center, Juniper is providing the needed horsepower and intelligence to set the bar for agility, assurance and automation. Customers will now be able to even more effectively build and reap the benefits of AI applications while saving substantial time and money.
What's Next?
Twenty-eight years ago, we made a bet.
The internet was at a near breaking point as early telecom companies were seeing network traffic double every three months, and the underlying infrastructure simply couldn't keep up. Back then, Juniper wagered that a new type of router with purpose-built silicon, a modular design and an internet-scale OS would unlock the future of the internet.
We were right. And it's not an overstatement to say that the internet wouldn't be where it is today without Juniper.
Now, we're at another inflection point with AI. But it will be a bigger shift than the internet and change at that scale will require us to rethink how we approach networking.
Because I imagine a world where there simply are no more network trouble tickets. Anywhere.
No application downtime or performance issues. No network outages. No one even knows that there is a network operating behind the scenes. There isn't a single human touching the controls of the infrastructure. In fact, there are no controls or UI at all… there's no need because everything just works, and it works well. Across every single domain.
This is a vision we set out to achieve years ago and we've made real progress over the last several years. And now, we have all of the pieces in place with our AI-Native Networking Platform to bring it all together and fully realize this vision.