All Tech Field Day Events

Running Private AI Data Centers Gets Easy like the Public Cloud with Juniper Networks

In tech, no other buzzword has caught on the way as the word “convenient”. Vendors, over decades, have deployed it to describe products and solutions, spreading its appeal far and wide across the industry.

Into that came public cloud with its white-glove service and zero-housekeeping, deepening the enticement to an irresistible degree. For the first time, users could add things on the fly on rented infrastructure without bothering with the technicalities.

Since cloud, user convenience has become an obsession and a mantra for technology vendors.

Now as corporations are plowing money into building glitzy AI data centers, the technical struggle is getting real, leaving them yearning for the same frictionless quality that is the hallmark of public cloud.

The expectation has also gone viral among organizations repatriating workloads from public cloud to private data centers. In the last three years, more than 93% companies have engaged in cloud repatriation projects, making it evident that private infrastructures are making a comeback.

The work they have on their hands is markedly inconvenient. “The workloads, the applications, the use cases we are trying to drive out of modern data centers are pushing scale to a ridiculous level,” remarks Nick Davey, director of product management, at Juniper Networks. “We see complexity and connectivity like we’ve never had to provision before in data centers.”

Cloud companies have made billions of dollars by reducing the behind-the-scenes effort of building and managing colossal infrastructures down to the click of a button. Now the question that is on everybody’s mind is – can the cloud experience be replicated over to on-prem?

Ready Templates to Build Network Fabrics

Historically, networks have hand-wired by network engineers. But that process has been the source of innumerable problems. Errors and fails are rife in manual processes, besides it being a more difficult and time-consuming way of working.

Juniper Networks is working to replace the manual work with an easy, hands-off, cloud-like experience that does not get network engineers deep in the weeds of the infrastructure.

There are many great examples of products that are low-complexity and low-effort on Juniper Networks’ portfolio. The mission to make the job of building massive data center networks trivially easy, however, is a high aim. But the company has a history of amazing advances to fall back on. For example, the AI Juniper Validated Designs (AI JVD) is a great place to start for enterprises looking for a light and easy experience with planning and designing large-scale networks.

JVDs are standards-based fabric designs for data centers – pretested and validated network diagrams, configs, protocols and encapsulations – that make a quick work of deploying large data center networks. The designs tell you where a product fits, its targeted use cases, and the best practices.

“We took all of the science and all of the research that we’ve come up with in our labs, all the testing and qualification that we do around our products and our use cases, and we packed those into a set of validated designs,” Davey says.

Juniper Networks describes these as the “guide to deploy the most complex networks”.

“We included one more important thing and that’s the automation actually required to make these things spring out of the box and come to life,” said Davey.

JVDs help users take advantage of years of learned experience and put it to work. The readymade templates help bypass the usual complexities and risks of deploying network solutions at scale.

But by no means is it a one-size-fits-all solution. To make JVDs universally handy, Juniper Networks tailors the designs to a multitude of use cases, and separate them based on the network size. They call these T-shirt sizes.

“The dimensions or the physical hardware making up these AI data centers, all of that gets packed in as parameters into our automation so that you can tweak and tune your AI JVD to match your networks requirement.”

Automating the Manual

Apstra is Juniper Networks’ solution for multi-vendor fabric management. It is is an intent-based networking software that automates tasks from Day 0 through Day 2, making it easy to manage networks of any design and topology like cloud.

“Operators and application owners are very used to clicking a button and receiving the thing that they asked for. They don’t go into a protracted set of meetings, or open tickets. They just push a button or slide a credit card and get the thing they want.”

Apstra is developed to deliver that fuss-free click-button experience. Apstra covers all three bases of design, deployment and operations in networking. At a high level, it is the central point of control for the entire data center fabric. Think of it as having a cloud-like API to consume all physical resources in the data center, Davey says.

While designing Apstra, Juniper Networks has embedded the common principles of network engineering into the architecture. Apstra offloads all of the backend designing and planning works from the whiteboard over to a virtual instance where teams can design, tweak and pre-stage a network and get a complete preview of the model.

“Until you click “Build”, it doesn’t assume that you have hardware,” he says.

Once a mockup is approved, the team can assign physical resources to it, plugging them in one by one.

Davey highlights that the hardware could be Juniper QFX switches, or any other equipment from vendors that Juniper Networks supports.

Using zero-touch provisioning, all of the hardware are brought up into the configured fabric.

Through Day 2 and beyond, Asptra monitors and manages all assets making sure everything is tuned optimally and running in perfect condition.

“This is the day-to-day work that carries on for years on end.”

Apstra leverages a fleet of probes and monitoring tools for this. “We’ve built a set of probes and optimizations that monitor all of the various cues and flows in a network and can optionally tune that network based on its observed values.”

This is a vital function for operating and maintaining of large-scale AI networks. “In AI, the scale of deployments that we need to bring in terms of both complexity and number of fabrics, and the demands of the workloads, we don’t want to do this by hand.”

AI JVDs borrow concepts from Apstra and embed them in the reference architectures.

“It’s already baked into the blueprints. Load the JVD into Apstra and then you can start from that foundation and tweak and tune it to match what you want it to be.”

The designs work equally for all kinds of data center fabrics. “We’re pre-staging all of the configuration for a typical EVPN data center fabric, but we’re mixing in all the sugar and spice required to make it a data center that can carry AI workloads, that means all the sets of configurations that make a fabric reliable enough to run AI workloads.”

Apstra makes it really simple to visualize the network topology and get a one-glance view of all the assets. But it’s harder to effect changes in bulk.

“The ultimate consumable interface is where you can order up your infrastructure just the same way you would order any other IT service.”

To give Apstra users the same power and flexibility, Juniper Networks is building integrations with automation tools on top of it.

A Terraform provider is now available to automate and streamline provisioning. “We view Terraform as the power tool, the universal remote if you will. Its job is to let us do bulk operations at AI data center scale with the happy coincidence of meeting the expectation of our cloud users.”

This is now connected to a ServiceNow workflow. When a ServiceNow request is raised, it calls the Terraform projects that then connect to Asptra and instantiate the network type requested.

The Terraform provider for Apstra is up on GitHub. You can also scan the QR code shown in the presentation to go to the links directly.

Watch more Juniper Networks presentations from the Cloud Field Day event at the Tech Field Day website.

About the author

Sulagna Saha

Sulagna Saha is a writer at Gestalt IT where she covers all the latest in enterprise IT. She has written widely on miscellaneous topics. On she writes about the hottest technologies in Cloud, AI, Security and sundry.

A writer by day and reader by night, Sulagna can be found busy with a book or browsing through a bookstore in her free time. She also likes cooking fancy things on leisurely weekends. Traveling and movies are other things high on her list of passions. Sulagna works out of the Gestalt IT office in Hudson, Ohio.

Leave a Comment