Visions Differ on Moving Containers to Production
The development of an ecosystem around application containers and related lightweight virtualizations technologies may be gaining traction as promoters of the technology move beyond the hype and focus on actually transitioning the technology to production environments.
One manifestation of the critical mass being achieved by Linux containers are gatherings in which key players can get down to brass tacks on the technical, process and production implications of adopting container-based virtualization. With that in mind, container proponents gathered in Austin, Texas, recently to flesh out the application container architecture.
The "Container Days" event was sponsored by lightweight container advocate CoreOS along with Docker container automation and management specialist StackEngine and others. The initial hurdle many see is that the current container ecosystem has too many moving parts.
What is needed to overcome this fragmentation is a more agile approach to moving containers to production, proponents argue. In a blog post summarizing the action items of the container workshop, StackEngine CEO Bob Quillin warned of "mass confusion" in enterprise implementation of Docker containers.
"For most organizations, especially enterprises, there are simply too many puzzle pieces to fit together," Quillin said. "From where to start, to which app to convert, rewrite, or start anew, which distro to run, container platform to evaluate, or service discovery to use, not to mention networking, storage, build process integration, deployment, scheduling, orchestration, and so on – the process is just too complex."
Quillin noted that Docker has been mostly an open source, developer driven concept consisting of a growing array of "pluggable components that developers compose together into a production application." The problem, of course, is that Docker and other container variants have made little headway in production workloads.
Because the problem of so many moving parts has been difficult to solve, proponents acknowledge that many developers have simply shifted to a platform-as-a-service approach as a way to hide complexity and actually implement container technology. "If the market doesn’t start to pre-assemble the pieces for customers, we’ll likely see more and more move to a PaaS model that hides this underlying complexity from the end user," Quillin predicted.
Instead, Quillin advocated a "new breed of solution [that] combines the ease of developer deployment of a PaaS with the infrastructure control of an [infrastructure-as-a-service] like Amazon AWS."
Google and CoreOS earlier this week rolled out their Tectonic approach that combines the existing CoreOS software stack with Kubernetes, Google's container management system. Google-backed CoreOS and Docker are attempting to integrate parts of the stack to fill in the gaps and push containers to production.
Quillin took issue with elements of that approach, arguing instead that the container ecosystem needs "better pieces to the puzzle to make everything enterprise-grade and production ready, from Docker to containers to the associated compute, networking and storage optimizations."
What is needed, the StackEngine CEO argued, is less attention to "assembling the puzzle pieces" and greater focus on reducing customers' licensing and other costs while enabling them to deploy applications faster.
Whichever approach wins out, the creative tension between competing visions will help transform container technology from hype to production.
Organizers of the container technology event in Austin said they would reassemble in Boston on June 5-6.
