InformationWeek: Let me take the discussion off the cloud for a moment to ask, were you surprised when federal CIO Vivek Kundra resigned in May?
Kemp: Yes, I was. Working for the government is a very difficult job. Few appreciate how little you can innovate from within the federal government. It's an institution designed for stability, not innovation. That's not a bad thing. But it makes being a change agent very difficult. As you get closer to an election, everything you do gets harder. It's viewed through a political lens. What may be the right thing to do gets viewed as something that's helping someone get re-elected. That's particularly difficult for change agents.
InformationWeek: How did you find your own experience inside a government agency?
Kemp: It was a great experience. Everyone should take time to serve in government or the military. I wish more people in NASA could go out and try to be entrepreneurs. If people could jump around to different companies, they would be able to see different infrastructures. When I left NASA, I left with no patents or special licenses. Everything we did became open source code. Once I resigned, I downloaded OpenStack [cloud provisioning open source code] just like everybody else. Too many contractors write into the contract that, if they solve the problem, then they own the software that they developed, while being paid by the taxpayer. They then go and sell it to another agency. That code should be open source. I'm just excited that there's so much interest in [NASA and Rackspace-sponsored] OpenStack. I hope everyone looks back someday and sees OpenStack as a valuable spinoff from NASA.
InformationWeek: What about trying to promote cloud computing in the federal government in a recessionary budget climate?
Kemp: All the budget stuff was a huge distraction. A lot of cloud stuff requires investment in staff training and change management. But most agencies are working under continuing resolutions that dole out a few more months of money. We didn't have the procurement or contracts in place to build the [NASA Nebula] cloud. The small amounts of money injected inhibited that.
InformationWeek: Are you happy to be back in the private sector?
Kemp: I am at heart an entrepreneur who somehow ended up in NASA. This is my fourth startup. I haven't made any money on any [of the previous three]. I'm still interested in doing something significant and successful. At NASA, we created Google Moon [Google's topographical maps of the moon built from NASA data] and Google Mars. We created the Microsoft Worldwide Telescope [a virtual telescope looking at NASA planetary image data]. We were trying to make massive amounts of data available [in remote data centers]. It inspired me to work on cloud computing at NASA.
Now that I'm outside NASA, I might be able to move the government toward cloud computing faster by having influence in the private sector. Even if Nebula has no government customers, it may have influence [on future government cloud decisions].
InformationWeek: What will be inside the Nebula cloud appliance?Kemp: The appliance will be a cloud controller, an aggregation place for server images, Web interfaces, a command center. It will have all the APIs for the features of the cloud, the storage service, the security service, and the OpenStack software. It's a box at the top of the cabinet. In it, you'll have all the management and orchestration through an enhanced version of OpenStack. There's also a switching fabric so each low-cost server in the cabinet is connected to the appliance by two 10-Gbps Ethernet switching cards. The appliance provides connectivity to everything in the cabinet.
InformationWeek: So underneath the appliance, there's a stack of servers? The appliance is not a cloud compute server itself?
Kemp: One of our appliances can power up to 24 Dell or HP commodity servers. The customer can decide which low-cost servers he puts in the cabinet. The moment you connect the appliance's Ethernet port to a server, the server will power on and become a compute node in a massively scaleable, elastic compute cloud. It will be provided with a lot of links. It will be compatible with standard APIs, from Amazon or Rackspace [public clouds].
InformationWeek: Customers can make a lot of different decisions. How does the appliance know what servers it's dealing with?
Kemp: It will hook into the existing configuration management database, if there is one. It's a lot less disruptive to IT, if it does. [CMDBs perform periodic discovery of devices on the network, capturing their configurations.] By having it talk to the existing tools, it can find out the existing configurations that may have been set to meet compliance requirements and other criteria. Our focus will be on the enterprise integration points, the really popular tools. The appliance will also log the events in every server and aggregate all that data.
InformationWeek: What tools will it first integrate with?
Kemp: It will integrate with Zenoss [network monitoring], Splunk [system analytics], BMC Atrium [configuration management database], HP's OpenView and ArcSite.
InformationWeek: What about CA Unicenter or IBM Tivoli?
Kemp: Unicenter is on the list. Tivoli is on the list. I don't know why I spaced on them. They're at the top of the list.
InformationWeek: You're combining cloud servers with storage?
Kemp: We're unifying compute cycles and storage. There'll be no expensive, Fibre Channel products in there [a jibe at Cisco's Unified Computing System]. Customers may put a storage server in the cabinet. The appliance will very dynamically look at the entire rack. It will have storage spread throughout. We will provide high availability, Internet-scale storage, lot's of storage.
InformationWeek: Is the enterprise ready to build out private clouds?
Kemp:Enterprises have three options. They can use the public cloud. They can hire a team of rock star integrators and build a custom cloud with open source and cloud software. Or they can hire a number of companies who have been doing business for 20 years and are willing to integrate outdated technology at your expense. The end result will not be as innovative as what you see in the public cloud.
That's why we started Nebula. What we have to do is build a great customer experience. We won't be able to deliver that experience in software alone--you need the hardware. By taking the [appliance] hardware and locking it down, we can deliver a lot of innovation. Apple has proven that. We want to be the iPhone of the data center.
Security concerns give many companies pause as they consider migrating portions of their IT operations to cloud-based services. But you can stay safe in the cloud. In this Dark Reading Tech Center report, we explain the risks and guide you in setting appropriate cloud security policies, processes and controls. Read our report now. (Free registration required.)