The jury is still out on OpenStack. As an open source set of tools for private clouds, it offers interoperability, standardization and open source compatibility benefits. It’s no surprise many of our service provider customers use OpenStack to power their cloud offerings. But it’s a different story for large enterprises. A lack of OpenStack skills, implementation difficulties and do-it-yourself support have left all but the most dedicated enterprises taking a wait-and-see approach. Yet, expensive licenses from hypervisor vendors and a slew of new modern applications require a new cloud infrastructure.
What will enterprises do? I predict that in 2017 the combination of containers and orchestration software will leapfrog the adoption and use of OpenStack.
Before I detail my prediction, though, let’s start with a simple definition of a cloud so we’re on the same page. A cloud must have these three basic elements: 1) Some form of virtualization, whether via virtual machines or containers; 2) some level of automation; and 3) most importantly, some level of self-service for individuals to provision their own infrastructure.
This, of course, is what OpenStack promises. It provides a cloud catalog similar to AWS — a full stack that can be implemented on-premises for your own cloud. But getting all of this up and running has proven more challenging than many companies expected. In addition, so far the deployment of OpenStack has been much more a top-down process where the business made a concerted effort to adopt the platform as part of a cloud transformation effort. Contrast that with the surging popularity of Docker, which bears a grassroots-oriented pattern. Developers are implementing Docker as a better way to build, run and ship applications on their own laptops.
In fact, this bottoms-up adoption eerily parallels another tectonic shift in IT: server virtualization. VMware catalyzed the shift to x86 virtualization when it released Workstation in 1999, a grassroots technology installed on desktops and laptops. From there it spilled into the data center and created the largest architectural shift since client/server computing.
But for Docker and containers to leapfrog OpenStack, it has to be about more than just easy-to-install technology. It needs to be about orchestration. Just like VMware introduced vCenter as a way of managing VM-sprawl, enterprises need a way to manage the vast number of containers that can be created. Combining containers with orchestration tools provides the necessary foundation for managing a next-gen cloud.
There are three noteworthy options. Docker Data Center, which includes Swarm — Docker’s own orchestration tool — Kubernetes and Mesos. All three are open source. All three have tremendous support from DevOps and developer teams. And all three have been tried and tested in true internet-scale architectures:
- Docker is supported by AWS and Microsoft Azure
- Kubernetes was born out of Google
- Mesos is used by Twitter, Airbnb and Apple
In short, containers shift development power to where it belongs: developers. Likewise, orchestration tools shift production responsibility to where it belongs: DevOps and IT ops. As a result, enterprises can manage and scale containers as part of private and hybrid cloud architectures. Optimizing the technology for the respective skill sets overcomes the skills gap barrier that top-down technologies like OpenStack impose.
I’m not predicting the demise of OpenStack next year or that containers will take over the world. What I expect to happen is that we will see a tipping point where many enterprises that were nascent with OpenStack, or haven’t gotten started, will instead move to Docker, Kubernetes, Mesos and others as a more right-sized approach to operating clouds.
Avinash Lakshman is CEO and founder of Hedvig. He founded Hedvig in 2012 after co-inventing Dynamo while at Amazon (2004-2007) and Cassandra at Facebook (2007-2011).