Juniper’s Approach: The Spark for AI-Driven Networks?

junipers-approach-the-spark-for-ai-driven-networks

In the digital age, network management has grown increasingly complicated and at an almost breakneck speed. Next-generation networks promise legions of microservices and network management challenges that most experts believe can only be handled by some form of network management automation, be it via deterministic means or through AI.

On October 10 in Las Vegas, Juniper Networks gathered industry analysts to discuss the company’s vision for next-generation networks. In light of the continued slow pace of adoption of software-defined networks (SDN) and network automation, it was interesting that the theme of the discussions was more about people and processes and less about cutting-edge technology. In essence, Juniper Networks feels that vendor solutions are not matching up well to where they have to run and who is supposed to run them. Today’s network automation push assumes that network engineers will think and are trained like software engineers.

Constructing a Unified Architecture

The company’s philosophy seems to come from the hyperscaler playbook, emphasizing open source simplicity and scalability as a key to delivering the complex services that are in demand from today’s and tomorrow’s networks. Bikash Koley, Juniper’s Executive Vice President (EVP) and Chief Technology Officer (CTO), who left Google in August 2017 where he was head of network architecture, said “Quite simply, you cannot automate a mess and you cannot automate well across many platforms siloed in each network domain or blindly stacked upon each other… There are workflows and intents common across many places in the network and we get economies of engineering scale and reuse through common platforms and specifications.” Juniper has constructed a unified architecture to underpin its software-defined products and has created processes, testing, and training to help network engineers improve how they run operations.

Koley is an evangelist for Site Reliability Engineering (SRE), which he sees as the key to operations success. SRE is the concept that engineering can bridge development and operations via metrics of reliability. SRE is engineering applied to operations, plus continuous improvement by automation. When you apply SRE to network operations, you get Network Reliability Engineering (NRE). According to Koley, “NRE concentrates on engineering as the approach to automation and the ops-familiar truth that availability is prerequisite to any other measure of success (including speed).” Juniper has fully embraced NRE and is spending time and resources promoting a five-step NRE process to automated network operations.

Improving the Process of End-to-End Engineering

So, what does all of this mean for the slick and shiny automation and AI-driven network operations solutions that are making their way into the marketplace? Isn’t AI supposed to usher in the era of intent-based network operations, where an engineer does not have to build, test, and release every network action, but instead can simply tell the system to do so? In the end, that vision of intent-based network management should remain, perhaps along with widespread adoption of a better process of end-to-end engineering, from development to operations, accelerating the market for automation and AI-driven network operations solutions. Why? Because end-to-end engineering removes the process and people messes that out-of-the-box automation solutions cannot.

Comments are closed.