Consistency may be the “hobgoblin of little minds,” but when it comes to automating network operations, maintaining a predictable, rinse-and-repeat approach is the name of the game.
The “learn by doing” platform developed by NRE Labs helps network engineers unpack and implement new automation and NRE tools that can help meet those goals—quickly.
“The whole value proposition of the software we’ve built is that it offers users the best of both worlds,” says Matthew Oswalt, product marketing manager for NRE Labs. “In addition to providing a full featured network testing environment, we offer the usability of a (how-to) blog post by taking the grunt work out of setting up an environment. We spin it up on demand, in real time, for each user.”
Launched in 2018, NRE Labs is an open source, collaborative project that supports a web-based, on-demand curriculum focused on the tools of automation. While the service is the brainchild of Juniper Networks, NRE Labs remains both brand-agnostic and free to users. The goal, says Matthew, is to foster a community that is built by engineers for other engineers, and to lower the barrier to learning new tools.
“A lot of network engineers out there are looking for that next step. We want to serve that community—and not in a Juniper-specific way, but rather in a way that helps people accelerate their learning across a wide range of skills and technologies.”
You Dropped a Bomb on Me
By logging onto NRE Labs, engineers (or anyone with an interest) can experiment almost instantly with new automation platforms using live terminal access to their own network devices and Linux systems.
The no-track, browser-based platform includes a curriculum that’s broken down into Fundamentals, Tools, and Workflows, covering topics like “Introduction to YAML,” “Working with REST APIs,” and “Junos Automation with PyEZ.” Lessons are comprised of real-life scenarios and parsed into short steps in a live lab environment.
“Typically, when you want to teach someone automation, you send them to a bootcamp or GitHub read-me page, and they have to figure it out for themselves,” Matthew says. “We decided to build software in the browser that allows users to avoid the minutiae of setting up an environment — instead, they just go to a page and start interacting with the technology immediately.”
Along with live labs for tinkering, NRE Labs provides resources to help map out individual learning paths. This includes documentation collections (categorized by contributing organization, level of difficulty, and consultancy) as well as Antidote, a “community initiative to democratize interactive, dependency-free learning.”
“Antidote is the software platform that powers NRE Labs. You can think of it as the underlying technology that makes the whole thing possible—and NRE Labs is a specific curriculum we deploy on top of it,” Matthew says.
The goal is to create a sustainable, supportive automation-driven community and a go-to tool that empowers network engineers looking for answers on the fly.
The Goldilocks Principle
Staying on point with the mission has come with its own unique set of technical challenges for NRE Labs.
“We run Kubernetes, but a lot of the software that powers the lessons requires Virtual Machines to simulate network devices from Juniper, Cisco and others. Because these are virtual appliances, they don’t run natively in containers,” Matt says.
With this requirement, NRE Labs originally partnered with Google Compute Platform (GCP), which has a feature to enable hardware-accelerated virtualization. This helped mitigate some of the performance penalty that arises from nested virtualization, but it was still a problem.
“Adding layers of virtualization was both cumbersome and expensive. That’s what got us thinking that maybe we should run NRE Labs on bare metal, so we’d have only one layer of virtualization,” Matthew says.
NRE Labs also has a sense of humor about turning outdated methods of network training upside down.
In certain engineering circles, the Goldilocks Principles states that a solution “must fall within certain margins, as opposed to reaching extremes.” While other cloud providers promised bigger, better, and more, only Equinix's bare metal cloud was able to get it “just right” for NRE Labs.
“Architecturally, we don’t need that much compute capacity. Because we clean up lessons when a user is inactive, we don’t require a huge amount of RAM or a zillion cores,” Matthew says. On the other hand, faster core speed definitely moves the needle for the real-time experience that NRE Labs users are looking for.
Equinix’s c1.small config fit the bill perfectly. “Equinix’s c1.small offers a much faster core speed than any of the VMs we can get with AWS or Google, and as a bonus we’re not paying for hundreds of gigabytes of RAM that we don’t need,” Matthew says. “That, along with a strong API and Terraform provider, made building on Equinix Metal an easy choice.”
Equinix’s right-sized bare metal solution with no trade-offs on performance proved the perfect fit.