The Self-Driving Network by Juniper


Juniper Networks, a Sunnyvale, California-based IT networking company, is making strides with its self-driving network with its recent announcement of Juniper Bots, a series of new applications that simplify network operations by translating intent into automated workflows. On its website, Juniper advertises the Self-Driving Network as, “An autonomous network that’s programmed to independently carry out your intentions, the Self-Driving Network eliminates the complex programming and management tasks required today to run your network.”

Juniper has been investing in artificial intelligence to develop its autonomous network, the goal of which is that minimal human intervention will be required to manage the network, predicting and solving performance problems before users are affected, leading to a drop in operating costs. Juniper also says that security and reliability will also improve with its Self-Driving Network.

Network operators traditionally interact with networks through a type of human-computer interface known as command line interfaces (CLIs), in which the user issues commands to the program in the form of successive lines of text. This is a labor-intensive process that involves lengthy design and configuration validation, and these complex programming languages that require highly skilled developers can be the cause of bottlenecks.

Network automation has gained traction as the alternative way to manage network operations. At its simplest, it means that it automates CLIs, but it can also involve script-level automation, intelligent network control and at its highest level, translate network administrators’ intent via policy.

Juniper Networks says that the essential underlying technology behind network automation (that which manages human-to-machine interaction) is still in its early days. In a recent study, Juniper found that the two biggest barriers to customers and partners adopting more network automation was, “the lack of an integrated end-to-end solution” and “the lack of internal education and skillsets”.

Juniper says it has “solved” these problems with its Juniper Bots, which “transcend complex machine-focused automation tools”. The bots analyse current network conditions and take action to correct any problems they diagnose. The bots or new applications translate human intent into automated workflows across network infrastructure, and cange the way that network operators interact with the network.

The new bots include the AppFormix Health Bot, the Contrail PeerBot, and the Contrail TestBot. Donyel Jones-Williams, Juniper’s ‎Director of Product Marketing Management, said, “We’re announcing three bots, but there’s a whole library of bots we’re going to introduce”.

In Juniper’s new podcast series, Jason Venner, VP of Architecture and Technical Marketing Contrail and Cloud at Juniper, said that there is “a massive pressure to innovate faster and more cheaply and deliver higher quality to its users”, describing it as the antithesis to the old rule of thumb, “give me two: faster, cheaper or better”. People want it all. The move to hybrid cloud, DevOps, continuous delivery, and machine learning is leading to data driven companies that are able to “recognize opportunities in the marketplace in near real time and deliver changes to their customers almost immediately”. Giant data companies like Facebook, Amazon and AirBnB are embracing this digital transformation, making it their core practice. They are demanding new technologies, like network automation, which allow them to keep up at the rapid pace of the Internet.

In order to stay up-to-speed within the present day digital business world, in which success is measured on changes per second from changes per month or per year, data driven companies are usually running on multiple data centers and a mix of infrastructure. Venner frames the challenge as, “How do you deliver an underlying network that is API driven that can handle these high rates of change and adapt dynamically to changes in your application, real world problems with your infrastructure, unexpected events and deal with bad actors, all while delivering excellent customer experience across multiple environments?”

Venner says these have to be delivered by automatic intelligence, as manual network operations cannot keep up with the pace of change necessary. Juniper aims to deliver high up-time at a low cost under high rates of change at enormous scale by automating the cloud and data center, and employing smart hardware, machine learning and SDN products to deliver API-driven networking.

Juniper’s new bots are the latest stage of this continuous digital transformation that the company is undertaking. The Contrail PeerBot automates the process of network peering by managing Border Gateway Protocol (BGP) routing and complex policies. The Contrail TestBot involves a DevOps approach to continuous integration/continuous deployment (CI/CD) and the AppFormix HealthBot is a machine learning health tracker for the network, which draws upon Juniper’s AppFormix to gather real-time network data and use it for trouble-shooting problems that arise.

Juniper purchased the startup AppFormix in 2016, which uses machine learning to provide telco cloud operators real-time and historic monitoring and performance optimization; in order to merge it with its own Contrail technology. The resulting bot is the AppFormix HealthBot. “This is the collector that sits on the other side of the Telemetry Interface,” said Jones-Williams. “We’re taking all this data, using machine intelligence, and checking the health.”

Juniper also announced enhancements to its Juniper Extension Toolkit (JET) to enable developers to create apps with direct data plane access on Juniper’s vMX and MX Series 3D Universal Edge Routers.

Scroll to Top