Did OpenStack Let VMware Into The Henhouse?

Still, this process, like any other, bears watching. To attract independent developers, the project must remain above the interests of any one or any small group of vendors, no matter what they contribute, and must not become too much the creature of big contributors. The data on code contributions suggest that with its latest OpenStack release, it's meeting that goal.

Rackspace was still the leading contributor to the Folsom release of OpenStack at the end of September, according to open source project tracking site Bitergia, but the data illustrates its declining role compared to the previous Essex release: it had 49 committers managing 973 commits to Folsom; six months earlier, it had 52 committers managing 2,275 commits to Essex.

The five other top contributors, showing the range of active code writers, were as follows: Red Hat, with 861; Nebula (ex-NASA CTO Chris Kemp's firm), with 464; HP, with 179; Information Sciences lnstitute at the University of Southern California, with 144; and CloudScaling, with 140. This is a healthy mix that shows OpenStack is not anyone's pet project.

Although OpenStack has been criticized as a loosely coupled set of code modules, Tucker said, "I think that's one of the project's strengths." Its working groups on Nova compute, Swift storage, Keystone identity management, Glance image management, Horizon reporting dashboard, and Quantum virtual networking will proceed independently and develop their areas of function as rapidly as possible while still maintaining compatibility in releases.

OpenStack shows vigor in other ways, too. Nimbula just became a member of the project, even though Nimbula, with early Amazon architects Chris Pinkham and Willem van Bijon as its founders, could have made a bid to go it alone with its own approach.

Cisco just released its Cisco Edition version of OpenStack, with additions in monitoring and management engineered by Cisco.

Red Hat CTO Brian Stevens said Red Hat will soon offer its own version of OpenStack, based on the more tested Essex release of OpenStack, not the Sept. 27 release of Folsom. Like other Red Hat products, it will be available on a subscription basis; no date has been set for release.

There are two major areas of debate within the young organization. Since many members have set their sights on competing with VMware for the operation of the enterprise cloud, some resisted the Aug. 28 acceptance of VMware as a gold board member. Did the fox buy its way into the henhouse? Board member Boris Renski, co-founder of the OpenStack consulting firm of Mirantis, says admitting VMware was a mistake.

In an interview, board chairman Alan Clark disagreed, calling VMware's joining OpenStack "a very positive move."

It's clear that VMware won't back off its long list of proprietary products, but it will try to ensure there's a way of interacting with a broader community of cloud users through the Quantum virtual network management module.

For the OpenStack board to have turned down VMware would have been cutting off its nose to spite its face. Open source projects are supposed to be open to contributors, and Nicira was a primary contributor to the OpenStack Quantum working group before the firm was purchased by VMware. Nicira will remain so after the acquisition, as signaled by VMware's gold membership. Yes, that puts the fox in the henhouse, but with so many watchdogs circling, it's likely to remain on its best behavior.

But of course, now VMware also has a better view into what's going on with KVM, a rival hypervisor recognized in OpenStack, now owned by Red Hat. OpenStack currently represents a neutral zone among competing vendors; everyone must take a share of responsibility in keeping the OpenStack cloud playing field level. And VMware has a stated interest in giving its customers the means of connecting to additional cloud providers beyond the immediate VMware fold.

Xen, an open source hypervisor project, offers a potential flag on how not to do things. I once harbored doubts that an earlier project, the Xen open source effort, would be open to all contributors on a somewhat equal footing. It appeared to me to be dominated by the interests of IBM, Sun, Oracle, and other large vendors that clustered eagerly around its launch, all of them wanting a vehicle with which to compete with VMware.

XenSource, the company formed to back Xen, eventually disappeared inside Citrix Systems, and Citrix Server and Oracle VM based on Xen continue to struggle to compete with VMware in the data center. With the emergence of KVM, independent open source developers who had been attracted to the project drifted away. And Xen's attempts to get its hypervisor supported in the Linux kernel ran into delays and repeated obstacles, not all of Xen's own making, but nevertheless resulting in friction with the Linux kernel development process.

We can pronounce that open source effort as something less than a smashing success. It still lacks the critical mass to become a dominant hypervisor, and KVM has gained broader support among independent developers, a critical group that broadens the general use of a piece of code.

OpenStack seeks to create a different opportunity, one where many competing interests create a viable, private cloud alternative, with service counterparts in the public arena. But time is short; VMware is already there.

There will be little room for infighting, enthusiasm-dampening big-vendor dominance or other hindrances to a full-fledged open source project driven by OpenStack's new governance. OpenStack needs all the initiative and imagination it can command, from as many developers as possible, to push the future of cloud computing forward.