OpenStack is many things, but a runaway success it is not. Despite a community that measures in the thousands, Gartner still counts OpenStack deployments in the hundreds -- on a good day.
This could change. OpenStack might, as Randy Bias urged the OpenStack faithful in his annual State of the Stack address, start streamlining development because "OpenStack is at risk of collapsing under its own weight."
But if you're a hyperfocused company like Red Hat, sinking even more resources in OpenStack development might not be the smart bet.
For years Red Hat has been trying to move beyond its founding mantra -- kill Sun and pricey, proprietary Unix -- to a new, but similar, mantra: Kill VMware and pricey, proprietary virtualization. Red Hat's embrace of OpenStack, an open source framework for managing virtualized data centers at scale (most often using Red Hat's KVM hypervisor), was a crucial part of this quest.
But OpenStack's future looks hazy at this point. So over the past year, Red Hat has jumped on the container bandwagon, embracing Docker, contributing code to Kubernetes for container management, and creating its own stripped-down Linux (Project Atomic) to compete with Docker-friendly CoreOS. Going forward, this is a better line of attack on VMware, which, despite recent efforts, will likely have a tough time transitioning to the brave new world of containers.
The OpenStack diversion
OpenStack has been a distraction for Red Hat, one that it seems to be finally shaking. Though the open source leader once dominated OpenStack development, this hasn't been true since the Icehouse release, when Red Hat accounted for 18 percent of all code commits (Mirantis was second at 13 percent, and HP tied for third with 12 percent). Since then, Red Hat's contributions have fallen off:
Juno: HP (18 percent), Red Hat (17 percent)
Kilo: HP (18 percent), Red Hat (16 percent)
Liberty: HP (19 percent), Red Hat (15 percent)
I'm sure Red Hat will dispute this characterization, and indeed, code contributions are an imperfect measure of commitment to a project. Red Hat's volume of contributions could be rising even as its percentage of total contributions falls.
Plus, as Kelsey Hightower rightly points out, there is far more to "contribution" than code. Documentation and other things don't show up in the stats above, but they matter a great deal.
Beyond code contributions, there are the public statements from Red Hat indicating it's all in on OpenStack. For example, while acknowledging that OpenStack users are early in their deployments of the cloud technology, Red Hat CEO Jim Whitehurst told investors, "We intend to be the vendor of choice for production deployments of OpenStack."
Maybe so. But despite Red Hat spending years pushing OpenStack hard, the company still saw only three of its top 30 deals include OpenStack last quarter, according to CFO Charlie Peters. That's not terrible, but it's not "OpenStack will drive huge revenue gains for us" either.
But containers might.
The Docker revolution comes to Red Hat
On its latest earnings call, Red Hat declared "phenomenal interest" in Project Atomic, Red Hat's stripped-down operating system designed to run applications in Docker containers. The company also pointed to OpenShift as benefiting from the container revolution.
But wait, there's more.
RedMonk analyst Fintan Ryan has pulled the data on Kubernetes contributions, an orchestration engine used to manage Docker containers. Red Hat trails only Kubernetes' founder, Google:
Kubernetes does one thing really, really well: manage containers. For enterprises that are turned on by containers' potential to upend traditional, expensive virtualization, Kubernetes offers an exciting option -- especially since their Linux vendor is now standing behind it.
I'm not suggesting that Red Hat will withdraw its support from OpenStack entirely. After all, Kubernetes and OpenStack solve different problems, and there's no question containers and conventional virtualization will coexist for some time. Plus, OpenStack has made its own efforts to integrate Kubernetes container management into its vast cloud management framework.
But I'd argue that to win against VMware, Red Hat needs Kubernetes more than it needs OpenStack. The "private cloud" market has been stuck in second gear for a long time, but few trends have ripped across the industry as quickly as Docker containers, mainly because they enable developers to package applications and deploy them in Linux containers with ease -- and without asking ops for help or permission.
Red Hat needs to double down on Kubernetes and containers. Yes, it's going to be a while before enterprises start deploying containers in production, but once they see the vastly improved resource utilization for themselves, they may never go back to conventional virtualization.
Of course, VMware has joined the container parade as well. It has its own micro Linux, Photon, to compete with Project Atomic and CoreOS, and it's contributing to Kubernetes. Very likely we can expect more container announcements at VMworld later this month.
But containers have been a developer-driven phenomenon, and VMware is a company with ops in its DNA, not to mention that pushing too far into the largely open source world of containers may have a disruptive effect on revenue.
Red Hat is several steps ahead of VMware in the container arms race. Plus, by virtue of its OpenShift PaaS alone, Red Hat enjoys a closer relationship with developers. The more resources it diverts away from OpenStack and toward container management, the better the prospects of beating VMware over the long haul.