Birth of a Hypervisor That Unleashed the Hybrid Multicloud Era

The Nutanix Acropolis Hypervisor (AHV) is in the limelight as CIOs and IT leaders across industries look for alternative webscale hypervisors and enterprise IT software following Broadcom’s acquisition of VMware*. The ten-year-old server virtualization software is a viable path for IT teams eager to migrate away from or counterbalance their reliance on VMware products. For others, AHV has been a bedrock of their software-defined, hybrid multicloud IT operations, which allow them to run and manage applications and data almost anywhere between private data centers and public cloud services. 

AHV was born of humble yet ambitious beginnings, just a few short years after startup Nutanix hit the scene with its pioneering hyperconverged infrastructure (HCI). The impetus for turning some bespoke operational code into a viable hypervisor product came from a simple customer question: “Where’s your user interface?” 

That’s what a customer in Japan asked Mike Cui, then a senior engineer at Nutanix, when he traveled to the organization in 2013 to set up a Kernel-based Virtual Machine (KVM) hypervisor. Initially, Cui was confused by the question.

“I thought the customer made the purchase because they were already familiar with KVM,” recalled Cui, who left Nutanix in 2019 to become a principal software engineer at AWS before his current role as senior staff software engineer at Google. “But he said: No, they knew it was an alternative hypervisor, but thought we had a GUI for it.”

The night before Cui flew back home, he learned that Nutanix had cut a multimillion-dollar deal with a government agency, a number that dwarfed the company’s average quarterly revenue at the time. The agency had only ever run bare-metal workloads, and leaders wanted to virtualize for the first time on Nutanix infrastructure, also with KVM. But Cui suspected that the agency would have demands similar to those of the Japanese customer. This new deal prompted Cui and the team to invest in creating a better interface between the KVM and storage.

“I came back, and I said, ‘We need to build something with a better storage protocol,’” Cui told The Forecast in a Spring 2024 interview.

“We needed to really solve storage. There were a lot of unknowns, but the deal was done, and the revenues were there. I knew we just had to make it work.”

Shortly thereafter, Nutanix hired Gregory Smith as a software engineer. Picked initially to work on storage innovation, Smith was roped into the AHV team to solve storage and management for KVM. This work shaped the Nutanix Acropolis Hypervisor (AHV) in 2014.

“It was a really scrappy solution at first,” Smith said, turning to Cui with a chuckle. Smith left Nutanix a few years after Cui and is a staff software engineer at Gro Intelligence.

Even the product name was a bit whimsical. Cui and Smith, who are friends outside of work, had taken a trip to Greece together the previous summer, and they code-named their beta solution after the Acropolis of Athens, but they never expected the moniker to stick.

“We named AHV after the birthplace of democracy,” Smith said. “We felt that we were democratizing the hypervisor.”

Like the name, though, the technology proved more durable than Cui and Smith initially expected. It brought simplicity, scalability, and management to a corner of the tech industry that had come to feel walled off, even to some IT professionals.

Let’s Build It Ourselves

Cui remembered how the startup was gaining traction in the industry for its innovative hyperconverged infrastructure. This, along with competition stirring in the industry at the time, motivated them to fully develop AHV in a generally available product.

“Before, there were some internal naysayers about what we were doing,” Cui said. “Now, there were fewer of them. That was a wake-up call.”

The first customer for AHV was Nutanix itself. 

“We not only built it ourselves, we also built it for ourselves,” said Cui.

Nutanix quickly moved all of its development and test environments to run on AHV. It was a bold move, as testing code was core to its existing and future products.

“We were constantly pushing it and trying to break it before our customers had the opportunity to,” Smith said. “We realized there were features we wished that we had before we ever got a request from a customer. That was really what drove most of the development in the early days.”

Steve Poitras, who was a principal solutions architect at Nutanix when AHV was created and later became chief architect before moving to DevRev, Inc., told The Forecast that this bootstrapping approach reflected the philosophy that drew him to join Nutanix in those early days.

During his hiring interview with co-founder and then-CEO Dheeraj Pandey, Poitras remembers seeing the “horrible” office carpet with cables strewn around the floor, like a quintessential engineering-driven startup. Above all, he saw Pandey had a grand vision with the technical chops to match. Poitras left his cushy job immediately to join Nutanix.

“Dheeraj’s whole mentality is: ‘Hey, you know what? If there’s not a good option, let’s build it ourselves,” Poitras said. “It’s actually a really good way to push innovation.”

One thing that immediately set AHV apart in the marketplace was its simplicity. Competitive hypervisors had what Smith calls “nerd knobs” that allowed engineers to endlessly, meticulously tweak their systems.

“It led to a very complicated experience, especially for processes like upgrades,” Smith said. “We decided very deliberately to sidestep all of that by presenting a very simple interface. Both Mike and I had been around long enough to have worked with interfaces that were more complicated, and it often meant that things became hard to predict and hard to quantify. Nobody wants that.”

Evolving the Industry

Poitras likens a hypervisor to a smartphone or a smart television allowing users to run and access multiple types of resources on a single piece of tech equipment.

“Back in the day, cell phones didn’t really have the concept of apps,” he said. “It was just phone calls and text messages.” 

However, they evolved to be mobile platforms for running applications and accessing the internet.

Similarly, hypervisors allow users to run multiple virtual machines on one physical server. The technology paved the path for the rise of virtualization, which has allowed organizations to shrink their data centers, conserve energy and reduce maintenance burdens. Virtualization, in turn, led to the public cloud and hybrid cloud environments that have become commonplace today, giving IT departments the ability to improve disaster recovery, increase visibility and portability of workloads, and dynamically scale resources up and down or across private and public clouds in response to changing demands.

“With physical infrastructure, it’s very difficult to manage, provision and automate,” Poitras said. “There’s a lot of friction when it comes to moving resources around. It’s a very high-overhead, manual, intensive way of doing things. The abstraction of a hypervisor allows you to live-migrate virtual machines. That opens up a whole plethora of possibilities.”

AHV has evolved dramatically over the past decade. The simplicity of AHV’s interface attracted many customers, and over time, beneath that simplicity, security, reliability and performance kept improving. Poitras remembered how the early presence of federal customers underscored the need for the AHV team to lock down security.

“Of all the hypervisors, AHV was probably the most secure, just because we had full control over it,” he said. “With others, we couldn’t get really detailed and change source code or core functionality for the hypervisor. But with AHV, we could. That’s where we built in a lot of these ways to mitigate threats and harden and make it more secure.”

Rather than relying on the underlying OS kernel to read and write data from hard drives, the team designed AHV to access data directly from userspace within the Acropolis Operating System (AOS). This allows AHV to bring better reliability.

“By moving out of the kernel we could treat storage like a network protocol, using networking concepts such as load balancing to achieve fast failover and high availability,” Cui said.

It also allowed AHV to scale up and leverage multiple threads for more parallelism in the storage data path.

“That was amazing to see because the performance became incredible,” Poitras said. “We could do 100-microsecond I/Os. Without that direct access, it was impossible.”

“What we’re doing, at the end of the day, is abstracting things as much as possible,” Poitras added. “AHV can manage itself, update itself. It does its job, and it does its job extremely well. It’s a hypervisor you don’t have to think about.”

They Built a Foundation for Hybrid Multicloud Innovation

Poitras lauds Cui and Smith as “brilliant individuals” whose work upended the way enterprises approached their compute infrastructure. Cui and Smith themselves, though, remain humble when recalling their efforts to make AHV.

“To be honest, I did not expect AHV to have this level of success,” Cui says. “At the time, we felt like if we just got something to work, that was already a huge accomplishment.”

Cui and Smith both said that Nutanix was able to build AHV partly due to other advances in the industry. Cui notes that the “work of multiple companies, over decades” brought virtualization to a point where the time was right for Nutanix to push the ball forward.

“We were in the right place at the right time,” Smith says. “We were able to take advantage of various things from the open-source world.” 

And then there were things happening in the industry as hyperconverged infrastructure began catching on. 

“If all of that hadn’t happened, we wouldn’t be shipping AHV as broadly as we are today.”

Initially, Cui didn’t think the work that led to AHV would do anything other than satisfy those early customers who were asking for a more intuitive user interface to manage their virtualization environments. It wasn’t clear that AHV would find large-scale traction outside of Nutanix’s own operations. But as work continued, he realized that this was a product whose time had come.

“It was through these gradual changes that AHV eventually became what it is,” Cui said.

“For me, one of the proudest moments was getting all of the developers up onto the dev cluster and then being able to migrate that to a different data center. That felt like: ‘Okay, we’ve hit the mission; this is what we were trying to build.’ Everything that came after that has just been the icing on the cake – an opportunity to push things a little bit further.”

Editor’s note: Learn more about Nutanix AHV and how hypervisors work, then compare Nutanix and VMware solutions and see a framework for migrating IT operations to Nutanix. 

Calvin Hennick is a contributing writer. His work appears in BizTech, Engineering Inc., The Boston Globe Magazine and elsewhere. He is also the author of Once More to the Rodeo: A Memoir. Follow him @CalvinHennick.

*Nutanix, Inc. is not affiliated with VMware by Broadcom or Broadcom. For additional information and important legal disclaimers, please go here.

© 2024 Nutanix, Inc. All rights reserved.