Are containers ready for prime time? This week we'll look at that question from two different points of view: RedHat, a container vendor that's trying to build the tools for container customers, and LinkedIn, a container customer that's betting on containers being the foundation of its future technology platform.
The vendor perspective
I recently talked to Lars Herrmann, general manager of Integrated Solutions at RedHat, about the state of Linux containers today. We agree that the No. 1 problem facing containers is education. The hype about containers continues at a breathtaking pace, but knowledge of how to effectively use containers lags sorely behind.
RedHat's been making a big bet on containers, and over the last couple years shifted its OpenShift solution from a PaaS platform into a container management platform. It hopes that the tools it's creating will make it easier to build -- at least initially -- greenfield new applications using containers.
There are a lot of security concerns around containers, Herrmann said. However, he believes Linux containers might be a much better way to do security, and points out that it enables an immutable software packaging system, unlike other options out there.
Finally, Herrmann pointed out the hybrid cloud opportunities that arise using OpenShift. With RedHat, you can use the same container management tools in the public cloud that you use in the private cloud.
So the container story told from the vendor perspective is focused on education, tooling, security, and hybrid cloud -- four points I believe are valid and interesting. But how do they match the concerns of a container customer?
The customer perspective
This week LinkedIn announced an internal container-based PaaS solution called LPS. Over the next two years, the company plans to move most of its mission-critical website onto it. Third party developers won't be getting access to LPS in the foreseeable future, but developers at LinkedIn will be able to deploy code faster and easier than before.
I talked to Steve Ihde, head of the LPS project that LinkedIn has been working on for the last year. The first thing I wanted to know was whether the company was using any of the big container managing platforms out there, including OpenShift, Cloud Foundry, Mesosphere, or Docker.
LinkedIn evaluated the big open-source options, Ihde said, but chose to forgo them all and instead build a solution itself on top of runC, a small sub-component of the full Docker container solution.
Clearly, many of these mature platforms had a lot of features baked in that LinkedIn had to either rebuild or omit. So what was the reason for not basing its platform on an existing platform?
The reason was not education, tooling, security, or hybrid. Instead, the first factor in LinkedIn's choice was self-reliance. Having to call a vendor at 4 a.m. on a Sunday if there was trouble with LinkedIn's website was considered not a manageable risk. It was critical that the container platform's operational knowledge and expertise remain in-house, Ihde said.
The second factor in the decision was migration. LinkedIn didn't want to build greenfield new applications using containers, it wanted to move its existing applications into containers over the next two years. LinkedIn had a choice to either re-architect all its applications to work with an existing platform like OpenShift, or build just one custom platform that minimized the cost of migration for all its applications.
When asked if LinkedIn would ever consider using public cloud for its mission critical website, Ihde's answer was a resounding no. The reason was cost -- cost of migration and ongoing operational costs. LinkedIn has no interest in a hybrid cloud because it has no interest in the public cloud.
What does this all mean?
If I had interviewed Netflix for this article, the perspective could have been very different. In some ways, LinkedIn and Netflix are almost polar opposites in terms of the conclusions they have drawn around public vs. private cloud. So which point of view is right -- or is hybrid really the answer?
Obviously, there is no one right answer. These stories demonstrate that Linux containers transcend the boundaries of the big public/private/hybrid debate. Everyone's beginning to adopt containers, no matter what kind of infrastructure runs underneath.
I also believe the concerns of container customers are not being fully addressed. One of the smartest decisions Docker has made was to subdivide the Docker ecosystem into small, self-contained components so it's easy for a company like LinkedIn to dip its toes in, one project at a time. Many of the larger container projects struggle to achieve mass adoption in large part because of the size of the commitment required, committing to so many architectural decisions at once.
Asked if, with hindsight, he would have made the same decisions a year ago, Ihde emphatically said yes. I believe that's a clear indicator there is still work to be done before containers are ready for mainstream adoption.
If you want to follow along in this often baffling journey to find out exactly when containers will be finally ready for you, sign up for bonus content and email alerts at ContainerMadness.com or follow me on Twitter @containermadnes (Twitter won't let me spell madness correctly which also makes me so mad).