buchachon - Fotolia
At the Nutanix .NEXT 2017 user conference, hyper-converged pioneer Nutanix took its next steps toward becoming...
an enterprise cloud vendor. Nutanix revealed a partnership with Google around its new Xi Cloud Services and Calm orchestration software and outlined other pieces of its cloud strategy.
"Hyper-convergence is just a pit stop on the journey," Nutanix CEO and founder Dheeraj Pandey said of the evolution.
We spoke with Pandey at the conference, and the Nutanix CEO elaborated on what the Google partnership can do for Nutanix, why the hybrid cloud is another form of convergence and how running a public company is like "a marathon of sprints."
What role will Google play in Nutanix Xi Cloud Services?
Dheeraj Pandey: It's about making hybrid invisible. How do you make hybrid cloud a true hybrid cloud so that it doesn't sound like converged infrastructure?
Because hybrid can be extremely hacked up, the way converged infrastructure was hacked up. We had a purist's view on what converged meant because it was all about software and running on homogenous x86 servers. It started on compute and storage, but then we added virtualization, we added networking and security.
So that journey has to continue. We're seeing that we now need to converge two different things, and they don't need to be hardware boxes. They happen to be data centers. One of them is owned by us, the other one is not owned by us. That's how Xi was born. Then the question was, how do we execute in Xi?
The Google partnership came along just at the right time. We said, rather than going it alone and learning about security and geographic distribution and dark fiber networks, how about we actually double down with a really good partner who understands these things really well? And [who] also has a very good handle on high-level [platform as a service] PaaS that they're working with developers on. And that will basically bring a cloud that is both PaaS and [infrastructure as a service] IaaS because just simply infrastructure by itself is interesting, but not complete.
What, specifically, will you do with Google technology?
Pandey: We will take Kubernetes, for example, and try to make it a hybrid cloud API. Now, what does that mean? Kubernetes right now is focused on containers. We will obviously make it run both in private and public clouds, but we'll also make it work between containers and virtual machines. You have to hybridize that as well. And Calm becomes that place.
Calm can be used to hybridize the experience between containers and virtual machines in private and public clouds.
If you can implement Xi inside Google Cloud Platform, then it's a win for both sides.
Will Google host Xi or will you maintain your own cloud, as well?
Pandey: Well, we're not exclusive with Google, so we'll continue to do our own things, as well. But we can actually learn what Google is very good at because there's an architecture thing here about whether we can just use the POPs [points of presence] to just backhaul everything to a large data center, or do we need or have smaller data centers in the metros? These cloud architecture discussions will clarify what you just asked.
You said you had a different take on converged infrastructure from the start. Do you see the early converged infrastructure vendors now sharing your vision?
Pandey: Of course. The thing is, they've come to hyper-convergence after four or five years, and the ball has moved yet again to hybrid clouds. Hyper-convergence is just a pit stop on the journey. It's really about software and operating systems. It's about being able to blend as many things as possible, and public cloud is one thing that cannot be a silo.
With everyone getting into hyper-converged, do you have to keep the ball moving to stay ahead?
Pandey: There's no finish line in the journey for companies that believe in innovation.
Dheeraj PandeyNutanix CEO and founder
You do the iPod, then you do the iPhone, then you do the iPad, then you do the iWatch, and you do the macOS; there's no finish line. Then you do the new Siri, then you need to do the car.
Microsoft went through this, too; you first went against Lotus, then you went against Novell, then you went against PC players, then you went against Oracle, and on and on and on.
At .NEXT last year, Nutanix was still waiting to go public. After a long wait, you finally completed your initial public offering last September. How has going public changed Nutanix and your job as Nutanix CEO?
Pandey: In many ways, it hasn't changed us, and in some ways, it has. We can't be confused about who we want to be. We are still a growth player. We are still about innovation. We are still about disrupting the incumbency, and that hasn't changed being private or being public.
Wall Street won't force us into something else. Only when something stops working because they don't see innovation, when they don't see modernization of research and development, will they say, 'look, if your strategy's not working, you better use my strategy.'
For a young public company, sometimes it only takes one bad quarter for that to happen.
Pandey: To me, Wall Street is like strict parents. If you do a good job, then they'll let you be, and they want you to be a better company every day. That's the way the board has been for this company the last seven or eight years. All they're saying is 'when you're building your company, make it a marathon of sprints. Every quarter is a sprint, so make it a marathon of sprints. I'm not going to get in the way of your marathon vision, but don't lose track of the sprint model.'
By the way, that's where development is headed. Agile development is about sprints. Engineers are impatient: I can't wait two years to release my code. Therefore, there are sprint models in Agile development. Every quarter of financial reporting, that scrum is that Agile development that we're actually doing in some sense. People bemoan quarterly milestones, but at the end of the day, that's what we're expecting our engineers to do.
What is your marathon vision as the Nutanix CEO?
Pandey: From very early on we have been clear and had the conviction for what the architect of computing needs to be. Many others bent over backwards with VMware. We didn't.
When everyone was building a tab in vCenter, we built Prism. We said that's the way we need to rearchitect the data center. It was bold and audacious. We don't have to bend over backwards. We have to have respectful discussions on strategy, and that's what we're having with Wall Street right now.
If our strategies work, if our research and development is monetizing and we are growing at the rate we're spending, then we'll be fine. If we start spending more than we grow, they'll ask questions.
On the business side, you must spend a lot of time managing relationships with server vendors. You have partnerships with all the major vendors or their resellers, yet they all sell hyper-converged appliances that compete with Nutanix.
Pandey: Apart from Dell EMC, which owns VMware at an arm's length, none of the other server companies have an operating system. They own a hardware company and they need an OS. The only OSes they can find are VMware, Microsoft and Nutanix.
As with most things in life, you focus on the overlapping interests and don't focus on the negative part, which is the conflict of interests. And that's how a community of nations operates. You can't be complete friends, you can't be complete enemies. You have your own interests. You focus on the overlaps.
You have no formal partnership with Cisco, and it sells a hyper-converged system, but resellers sell your software on its Unified Computing System. Can you see a formal partnership developing with Cisco?
Pandey: I never say never. Some of our large customers experience our software on Cisco hardware in production. So Cisco is now hearing from the grassroots, 'this thing is good and it's working,' and the more they listen to their customers, the higher the probability of us working together in an official partnership.
But I can't give you a date, I can't give you a time. We're in this negotiation process where the market forces will tell them, 'This is a good thing.'
Do you see Nutanix ever getting to the point where you only sell software on other vendors' hardware instead of on your own appliance?
Pandey: I don't think we'll ever abandon the appliance. That reference design keeps everybody honest in terms of how hardware and software comes together. It shows how they should behave, shows us the failure rates.
Without hardware, we could not build Xi. Xi is end to end; it's not just software running in the ether. It needs silicon. That's what our appliances give us. If we didn't know how to support all that, it would have been way harder to build Xi.
Hyper-converged users who don't buy Nutanix often say it's because Nutanix costs too much. How do you respond to potential customers scared off by price.
Pandey: I think people focus too much on hardware. They focus on Capex when they really should focus on working the whole body; what we call the cost of ownership. The hardware players want to provide cheap hardware, but then they charge a ton of money for services and support and things that come afterwards. We, on the other hand, believe a lot in automation [and] a lot in one-click experiences. That innovation comes upfront.
These days, we have so many different pricing models. We have subscription pricing, and Xi is another form of that. People don't have to look at it as Capex.
By the way, hardware vendors can't sell it that way unless you lease from them. But leasing is a hack actually. You really need a multi-tenant architecture, which is what the cloud provides.
But we don't talk about the cost of the hardware by itself. We talk about how many people will be required, how we'll provide you support for application, security, networking and migration. We've written so many tools that we give out for free; migration tools, testing tools, sizing tools. That stuff costs money to develop, and we don't charge anything for these tools.
Do you have a pricing model yet for Xi?
Pandey: It will come out [in the] next few months. But it definitely will be a consumption model-based approach. Not on cores because they don't see the cores; those are multi-tenant cores. It will be based on things they see, the consumed number of VMs [virtual machines], size of the VMs and storage.
Data centers can thrive when using hyper-converged infrastructure
Converged infrastructure and reference architecture are not the same
Storage is one winner as convergence meets the data center