OpenStack Is Not A Proprietary Cloud, Kemp Argues

Former NASA CTO and Nebula founder Chris Kemp says private clouds will need to be based on a flexible, general purpose set of open source code that can work with public clouds.
OpenStack is not a product. It's an interleaved stack of open source code that builds up layers of automated cloud operation from a single vision of how a cloud should perform, declared Chris Kemp, CEO of Nebula, a venture capital-backed startup betting on OpenStack.

With 226 code contributors to the project, working on Nova compute, Swift storage, Glance virtual machine image retrieval, and other processes, it is "the richest ecosystem in IT today," he added. Before he launched Nebula, Kemp was CTO of NASA and CIO of the NASA Ames Research Center.

"One of the most exciting things about OpenStack is the number of contributors," he said during a keynote address at the OpenStack Design Summit in San Francisco Thursday. But he also claimed that OpenStack has attracted a broad variety of backer companies, from startups to IBM and HP, from cloud service providers to component suppliers Intel and Calxeda. And it is ramping up faster than Linux did in its early days.

Nebula itself, with 50 employees, many of them engineers, is taking a leadership role in the development of OpenStack. According to one analysis, it is fifth in the number of employees contributing code. It ranks second, however, in providing code reviews that decide whether code is ready for testing and inclusion, a metric that has more to do with quality as opposed to quantity. It also ranks second as the company providing the most changed lines of code and most bug fixes going into OpenStack's new Essex release. In each case, it was second to Rackspace, which co-founded the OpenStack project with NASA.

[ Want to learn how Nebula wants to be like the Apple iPhone? See Nebula Wants To Be iPhone Of Data Center. ]

In his talk, Kemp said the OpenStack contributors work from a shared vision of what constitutes cloud computing, but the OpenStack code will be useful in building different types of clouds. "Amazon has made a set of decisions about how cheap is cheap (in infrastructure-as-a-service), how reliable is reliable, and how close to you is close enough" as it has built out its data centers. OpenStack users can make their own decisions about what constitutes a high availability cloud or a high performance cloud, and in each case they will be able to use OpenStack to build a cloud service with different properties.

"What we have the opportunity to do is build a platform that allows portability of applications (across other OpenStack-based services) to suppliers that are incredibly secure, incredibly high performance," or types of specialized services that will be the competitors of the future, Kemp said.

Amazon Web Services, the leading infrastructure-as-a-service provider, with revenue estimated to be approaching $1 billion, "is now a big battleship" that nimble OpenStack-based services may one day outmaneuver.

At the same time, other speakers at the conference, such as Mark Shuttleworth, CEO of Canonical, said OpenStack developers were conscious of AWS's EC2 architecture and APIs, and were determined to "offer a bridge to that other great cloud ..." and avoid irreconcilable incompatibilities. The project plans to be able to translate between its and AWS's APIs, so that an OpenStack user may form a hybrid operation with EC2. "It's a slightly controversial subject, AWS compatibility," he conceded at one point.

Kemp, in an interview before his talk, pointed out that Nebula has hired Tres Henry, the former Amazon principal engineering lead for AWS's Management Console. It has also hired Daniel Ford from Google, who led R&D on Google's approach to storage in a cloud system, as well as Matt Gambardella, who designed search engine servers for Microsoft.

Nebula is focused on producing a combination of hardware and software that will ship in racks to customers, get plugged in, and be ready to function as an integrated cloud service. It would offer virtualized and integrated servers, networking, and storage without configuration problems for the customer. It's still early in the process, but the first version of such converged infrastructure will be limited to running KVM virtual machines, with VMware's ESX Server, Microsoft Hyper-V, and Citrix Systems XenServer yet to be included among supported systems.

But Kemp said Nebula is building infrastructure for private clouds that will run "the next generation of cloud applications," not legacy applications, and that generation is coming fast.

He also called out a leading converged infrastructure product, Vblocks, from VCE, a consortium formed by VMware, Cisco, and EMC. "Vblocks is a collection of yesterday's most expensive technologies, installed by the industry's most expensive consultants. That's exactly the opposite of what we're doing," he said.

Nebula was formed in 2011 and named for the first federal infrastructure-as-a-service that Kemp helped create at NASA Ames. It's not yet announced a date when it will deliver its private cloud infrastructure product.

Find out how to move beyond server virtualization to build a more flexible, efficient data center in the new Private Cloud Blueprint issue of Network Computing. (Free registration required.)