Juniper brings AI bots to intent-based networks

The concept of intent-based networks has received a lot of attention from media and networking professionals since Cisco launched its “Network Intuitive” earlier this year. 

Cisco has certainly made the term “intent-based” a household term, but that wasn’t the first time I had heard a vendor talk about this vision. Years ago, I was at an event held by Juniper Networks where its founder and CTO at the time, Pradeep Sindhu, talked about the death of Moore’s Law and how that would drive us towards this thing called intent-based networking. 

Since then, Juniper has been somewhat quiet regarding its role in the evolution of the market, although it has been quite aggressive in the area of software-defined networking (SDN) through the evolution of its Contrail platform. Last week at its NXTWORK 2017 Customer event, Juniper tied its innovation with Contrail to the vision of a self-driving network with the release of Juniper Bots designed to translate intent into automated workflows.

It’s fair to say that all the great advancements the industry has seen in networking over the past few years — which includes the shift to software, increased adoption of white boxes, new operating systems, and the shift to software models — have enabled us to do so much more with our networks. But they have also increased the complexity of running a network.

The shift to the cloud has also raised the importance of the network, as we are now literally connecting everything to the network. Businesses have had to hire more people with new skill sets just to maintain the status quo.

Automation is something that network professionals seem more open to today than they did just a few years ago, but what to automate and how remains somewhat of a mystery. At the event, Juniper provided a data point from its research that found 43 percent of respondents said a lack of internal education and skills are preventing the use of network automation.

Juniper Bots help automate network tasks

Juniper Bots were designed to address that lack of skills. The solution consists of Contrail Intent Bots and AppFormix Analytics Bots, which facilitate automation by making it easier for people to interact with their network. Details of the Juniper Bots are as follows:

• Contrail PeerBot automates the process of network peering. This makes it easier to manage multiple Border Gateway Protocol (BGP) domains, simplifies policy enforcement and enables on-demand scaling.
• Contrail TestBot enables network professionals to shift to a DevOps approach for continuous integration/continuous deployment of network resources. The Bots can be used to automate auditing and provisioning modifications of the network. Humans simply can’t do these tasks fast enough to apply DevOps principals to the network.
• AppFormix HealthBot is like a Fitbit for the network. It uses machine learning to track the fitness and health of the network by leveraging AppFormix to collect real-time network data that can be used to discover new insights. The HealthBot translates the data into actionable information that can be used to troubleshoot and maintain the network.

Juniper Extension Toolkit updates

In addition to the Bots, Juniper also announced several updates to its Juniper Extension Toolkit (JET). The enhancements extend the management and control API framework to the data plane, allowing developers to create applications that directly interface with the data plane on Juniper’s vMX and MX Series 3D Universal Edge Routers.

Juniper’s plans for AI and machine learning

I had a chance to talk with Juniper’s current CTO, Bikash Koley, who came from Google and has been in the CTO role since July when Sindhu stepped down. I asked about Juniper’s plans for artificial intelligence (AI) and machine learning. He said most of AI that is in use today is what’s known as broad AI, like Amazon’s Alexa. This can tell people the weather, what time a meeting is and who is playing a certain song. This requires a broad knowledge base but doesn’t require much depth.

Zeus Kerravala

However, those types of tools don’t have the necessary information or smarts to run a network. That requires narrow AI where the data behind it is extremely deep but the functionality of the tool, like a Bot, is well defined and narrowly focused. Koley said over time we should expect Juniper to roll out more narrow AI Bots that solve the big problems plaguing network operations today. Long term, Juniper may venture into being a broad AI vendor, but the immediate focus is on simplifying network operations. 

Another good point Koley made was that making something simple actually requires a tremendous amount of engineering and innovation. Juniper will be applying its R&D efforts into masking the complexity of the network so its customers can do more, faster.  

Fulfilling on the vision of a self-driving network is still likely to be years away, and vendors will need to help its customers crawl, walk and then run to a fully automated environment. The Juniper Bots are a great starting point, as they let engineers leverage AI without putting the business at risk.