The Self-Driving Network™

Operationally efficient, solidly reliable, and remarkably adaptive.
In 2004, a handful of unmanned vehicles gathered at the starting gate of a lengthy course across the Mojave Desert—the inaugural DARPA Grand Challenge. The event sparked the historical start of a technological race to develop a viable self-driving car, a disruptive global movement that continues unabated today.

In much the same way, we’ve embarked on a journey toward a production-ready, economically feasible Self-Driving Network™. What we’ve learned from the progress of autonomous vehicles, we can apply to the evolution of network technology. The world is ready for autonomous networks. Advances in artificial intelligence, machine learning, and intent-driven networking have brought us to the threshold at which automation gives way to autonomy.

The networking community hungers for disruptive ideas to address the unsustainable economics of present-day networks. Today, operational complexity is increasing exponentially as traffic continues to explode and new devices proliferate. Rising operational costs and slower time to revenue are squeezing margins for traditional service providers, and limiting enterprise growth as well.

At Juniper, we see an answer to this problem taking shape in the form of the Self-Driving Network, a paradigm that will abolish operational complexity regardless of the type and volume of network traffic.

Beyond Predictive and Adaptive
The Self-Driving Network will take the heavy lifting out of the hands of IT staff. It will self-configure, monitor, manage, correct, defend, and analyze with little human intervention. It will be predictive and adaptive to its environment, optimizing and personalizing the experience for the end user and for the situation.

When traffic spikes occur on today’s networks, it’s difficult to distinguish a distributed DDoS attack from widespread downloading of the new Lady Gaga album. Using machine learning algorithms that interpret vast amounts of traffic behavior data, the Self-Driving Network will predict performance issues before users are affected. In this example, connections with algorithms that scrape Twitter feeds will confirm the hypothesis: Have hacking groups been threatening action against a particular enterprise? Or have fans been clamoring for Lady Gaga’s album in the weeks leading up to the spike? The Self-Driving Network will analyze and adapt accordingly, either shutting down ports to isolate the DDoS attack or adding bandwidth to accommodate the surge in album downloads.

Automation, Augmentation, and Then Autonomy
At Juniper, we envision the Self-Driving Network as the end state of a progressive journey that begins with automation and programmability and builds through the integration and advancement of four technology areas: telemetry, machine learning, intent-driven networking, and local and global awareness.

Automation and programmability have always been key considerations driving Juniper’s hardware and software development. Topology discovery, path computation, and path installation have already been automated. The Juniper Extension Toolkit (JET) for Junos® OS enables third-party applications to automate many other tasks. For example, bandwidth reservation is already responsive to traffic changes, but can we make it smarter? Can we automate service placement and motion? JET can help us move forward.

The Junos Telemetry Interface delivers high-frequency telemetry data to performance monitoring and optimization tools. But SNMP, pull-based telemetry, and naïve deep packet inspection are starting to show limitations. For the success of the Self-Driving Network, we need telemetry that’s based on push semantics and machine learning-based anomaly detection.

Machine learning uses algorithms to iteratively learn from data inputs. Instead of rule-based, static network programming (“If x happens, then do y”), machine learning algorithms recognize patterns in data, make predictions, and take appropriate actions without having to be programmed. This type of predictive analytics already exists in Juniper security products using heuristic algorithms. The more data fed into algorithms for networks, the smarter the networks will become—paving the way for intent-driven infrastructure.

Intent-driven networking also turns traditional if-x-then-y programming on its head. In this scenario, you tell the network what you want to accomplish, not exactly how to do it. Intent-driven capability exists on Juniper products today. SDN controllers such as Northstar and Contrail currently have built-in, high-level intent-driven virtual networking capabilities. And recent Juniper acquisition AppFormix uses intent-driven commands and machine learning technology to redefine telemetry and cloud operations management for software-defined infrastructure and application layers.

“Intent-driven networks are an important evolution enabling the network architect to describe what is needed, with network management software translating requirements into a design, and they will form an important part of the foundation of the Self-Driving Network. Adding intent-driven capabilities to SDN management and control software to facilitate automated network configuration and operation will be an imperative,” says Cliff Grossner, Ph.D., a senior research director and advisor for cloud and data center research at IHS Markit.

While local awareness or views of the network will remain essential, increased global awareness will help usher in efficiencies and optimization that would otherwise be missed in autonomous networks. Information correlation across local and global views, time, geography, network layers, and BGP peers will yield more insights, and those insights will make anticipating the network needs of customers and applications more accurate.

Augmented Intelligence, Not Artificial Intelligence
Advancements in automation and artificial intelligence technologies often evoke images of a robot uprising, at least in the workplace. While some mundane work will inevitably be handed off, technology has a habit of creating more jobs than it eliminates. The business world will always need people to oversee the network and design and monitor the robots. We’ll need people who understand how the network works to train the artificial intelligence, and experts to provide oversight and algorithmic tweaking.

The Self-Driving Network will only free network staff from the repetitive chores. IT will spend less time troubleshooting performance issues and running the network, and more time on strategic work and innovation that secures the business and drives it forward. As the Internet of Things gathers steam, these engineers will be in high demand to make sense of the deluge of incoming data. The Self-Driving Network will filter out the normal and allow IT to focus on the anomalous, the unexpected, and the dangerous.

Like the unmanned vehicles at the DARPA Grand Challenge, we’re at the starting point of a paradigm shift in the networking industry. Abstracting, simplifying, and obscuring network complexity is our new Grand Challenge, a course we’ll pioneer with the Self-Driving Network.

Juniper has invested substantially in automation and analytics development to set the foundation for autonomous networks. Automation and programmability are the first steps toward the Self-Driving Network, for service providers, cloud operators, and enterprises. Learn more about the drivers, technologies, and the journey toward autonomous networks in the resources for this article.