Born from NASA, Nebula Aims to “Disrupt and Democratize” Cloud Computing

8/11/11Follow @wroush

If you’re a young Silicon Valley entrepreneur, you’d probably give your Prius and your iPad, and maybe throw in your X-Men comic collection, to hear Andy Bechtolsheim ask you the question he posed to Chris Kemp this spring: “What’s the name of your company? I want to write you a check right now.”

If you don’t recall, Bechtolsheim is the Sun Microsystems co-founder and investor who wrote the $100,000 check that launched Google in 1998. The legend is that Sergey Brin and Larry Page had to incorporate under the name they’d picked so that they could actually cash it.

The 34-year-old Kemp, however, didn’t have a name ready. In fact, he didn’t even have a company—he was still a civil servant at the time of his meeting with Bechtolsheim, working as chief technology officer at NASA.

Kemp had led a project at NASA Ames Research Center in Mountain View, CA, to create open source software that would tie together the agency’s computing and storage resources into a private cloud computing environment similar to Amazon’s Elastic Compute Cloud (EC2) and Simple Storage Service (S3). The project was called Nebula, and together with other software components built by cloud hosting service Rackspace, it formed the core of an open-source software library called OpenStack. Kemp had conceived the project mainly in order to help NASA, the White House, and other government organizations get more efficient about the way they use computing resources. With cloud computing as hot as it is, though, Kemp had already heard from a few Silicon Valley investors telling urging him to take OpenStack commercial. He’d always brushed off the idea.

Chris Kemp, Nebula CEO

But when he finally met with Bechtolsheim, he says, “Something clicked. I thought, ‘if Andy Bechtolsheim is willing to slide a check across the table, perhaps I should consider leaving NASA.’”

Which he soon did. Kemp resigned from NASA in April and together with Devin Carlen, an engineer who’d worked on the Nebula project for both NASA and Rackspace, and Steve O’Hara, a telecom and semiconductor industry veteran, incorporated Fourth Paradigm Development—soon renamed Nebula. By May 1, the company had collected seed funding from a billionaires’ club that included not just Bechtolsheim but also David Cheriton and Ram Shriram, both early investors in Google alongside Bechtolsheim. The team had also signed venture term sheets with John Doerr at Kleiner Perkins Caufield & Byers and Peter Bell at Highland Capital. (The exact amount of money the startup raised hasn’t been disclosed, but when I asked Kemp “Are we talking about a $2 million round here or a $20 million round?” he indicated that it was closer to the latter.)

Having recruited this dream team of investors, Nebula announced its existence in late July at the O’Reilly Open Source Convention, better known as OSCON, in Portland, OR. It’s already hired 20 employees, about half working from its Palo Alto headquarters and the other half, all engineers, in Seattle. The company’s plan is to sell a hardware appliance—in essence, a very smart networking switch—that, once installed inside a company’s data center, will use OpenStack to automatically organize existing commodity servers and storage arrays into a private cloud infrastructure.

The attraction of the cloud model, of course, is that companies can save lots of money by spreading their big-data analytics tasks and other work flexibly across hundreds or thousands of computing nodes. The attraction of private clouds is that organizations that own them don’t have to pay Amazon’s rates, or risk putting their closely held financial information and other sensitive data on a public cloud. But while OpenStack is free, setting up and administering a private cloud is still tricky. “We’re saying that for people who don’t want to hire a big engineering team to administer OpenStack—if they really just want a turnkey private cloud—they can come to us,” Kemp says.

Nebula and its investors are betting that the movement Amazon started, and that Rackspace and many other “Infrastructure as a Service” providers are pushing forward, is destined to roll through most of the corporate computing world, the way minicomputers did in the 1970s and 1980s and client-server computing did in the 1990s. Nebula will “disrupt and democratize cloud computing,” Kleiner’s Doerr predicted in a statement released on the day of Nebula’s launch. He went on to say that Kemp’s team “has the unique expertise” to deliver the technology in a simple, scalable, low-cost manner. That remains to be seen—the company doesn’t expect to begin field trials of its appliance until the fourth quarter of this year, and there are plenty of other companies, including big ones like IBM and EMC, developing their own private-cloud appliances.

But Doerr is right about one thing—none of the competitors can count the pioneers of OpenStack as founders. More than 100 companies are already contributing to the OpenStack ecosystem, which seems likely to evolve into the open-source foundation for a broader cloud-computing movement, the same way open-source tools such as Linux and Apache undergirded the Web revolution.

But Linux probably wouldn’t have spread as fast as it did unless for-profit companies like Red Hat had come along to make it safer and easier for enterprises to use. So, is Nebula to OpenStack as Red Hat is to Linux? “It’s not a perfect analogy, but it’s an incredibly good analogy,” Kemp says.

Red Hat is in the service-and-support business, while Nebula intends to make money selling hardware. And Red Hat created many diverging versions of Linux, whereas Nebula’s hardware will always run the latest standard implementation of OpenStack, according to Kemp. But the analogy holds in that Nebula, following in Red Hat’s footsteps, will be able to add security, chargeback mechanisms, and other features that will make operating a private cloud feasible for companies that would never consider using a public cloud like Amazon’s.

Nebula’s isn’t building its first enterprise cloud appliance from scratch. Kemp says it will be built around a 10-gigabit Ethernet switch made by Arista Networks, the Santa Clara, CA-based high-speed networking company founded in 2005 by none other than Bechtolsheim. Since many companies are already in the process of upgrading the networks in their data centers from 100-megabit Ethernet to 10 gigabits, Kemp reasons, they’re likely to be in the market for switches already. “We realized that if we could take one of these top-of-rack switches that people are already upgrading anyway and attach an appliance that would auto-detect whatever was plugged in and determine if it is a supported hardware configuration, that would allow us to deliver this turnkey cloud provisioning feature,” he says.

That detail about “supported hardware” is important. Kemp says Nebula isn’t building a magic box that can turn any data center into a private cloud. The Nebula fabric controller—the main program that decides which computing nodes in a cloud to activate—will only work if the nodes themselves are commodity servers from companies like Dell and HP. “What we learned at NASA is that you can’t expect performance, availability, or security if you try to support every piece of hardware ever invented,” Kemp says. “The key to large-scale infrastructure is to homogenize things.”

Nebula’s business model will be pretty simple: customers will buy the appliances (for a price that’s yet to be announced) and subscribe to a support contract that will entitle them to quarterly upgrades to the latest version of OpenStack. Of course, the more appliances Nebula customers buy, the more switches Nebula will have to buy from Arista—which seems like a pretty good deal for Bechtolsheim.

I wondered aloud to Kemp whether that might have been Bechtolsheim’s motive for poaching him from NASA. “Andy didn’t poach me from NASA,” Kemp replied. “And [the Nebula appliance] is not just an OpenStack variant of an Arista switch—it’s more like OpenStack with a switch bolted on. But as a company that has invested trying to make the network faster and less costly, there is certainly some strategic alignment.”

Kemp says Nebula will continue to contribute to OpenStack. In fact, Carlen is leading a project called Dashboard, which will give OpenStack users a Web-based interface for monitoring and managing their private clouds. “The great thing about OpenStack is that it gives the market a way to contribute and innovate,” Kemp says. “By creating a reference implementation we are able to add security features that might be of interest to folks who aren’t Amazon’s target customers.”

Amazon itself, of course, is free to adopt all or parts of OpenStack for EC2 and S3—but even if it did, that wouldn’t change the economics driving enterprise interest in private clouds, in Kemp’s view. “What Facebook and Google and Microsoft have done with their own cloud systems is to say, ‘We can’t afford enterprise-class everything,’” he says. “Orders of magnitude cost savings can be achieved if the software is smart enough to take advantage of tens of thousands of commodity nodes. But Facebook and Google are the anomalies. We are trying to make it possible for the thousands of businesses that don’t have the expertise to build these scale-out systems to take advantage of all the great innovation that has happened with OpenStack.”

Wade Roush is a contributing editor at Xconomy. Follow @wroush

By posting a comment, you agree to our terms and conditions.