Eric Knorr
Contributing writer

Docker CEO: Our container goes anywhere

analysis
Apr 28, 201415 mins

The cloud era is in full swing, and Docker provides an answer to its greatest hazard: platform lock-in. An exclusive interview with Docker CEO Ben Golub

Few enterprise technologies have had such a quick ramp-up as Docker. Not yet in its 1.0 version, Docker is already wildly popular, because it enables you to package up any application in a lightweight, portable container and move it to any Linux server — a much more agile proposition than moving VMs around.

Recently, InfoWorld Executive Editor Doug Dineley, Senior Writer Serdar Yegulalp, and I sat down with Docker CEO Ben Golub to talk about the phenomenal early success of Docker and how it promises to disrupt enterprise application deployment.

[ Which freaking PaaS should you use? InfoWorld helps you decide. | Stay on top of the state of the cloud with the InfoWorld “Cloud Computing Deep Dive” special report and Cloud Computing Report newsletter. ]

InfoWorld: How does it feel to be the “it” technology?

Golub: I’ve been around for a long time — and have not been “it” like 22 times. First I was CEO of Gluster. I was CEO of Plaxo before that. Also eight years at VeriSign, a couple of years at Avid …

InfoWorld: You really have been around.

Golub: Yeah. The people at Docker are much closer to my kids’ ages than they are to mine. I feel very fortunate because we certainly happened upon a good solution.

InfoWorld: You stepped in after the technology was developed?

Golub: Yes. The company was first called dotCloud and was founded in 2011 by Solomon Hykes as a public PaaS, one of the first that handled multiple languages. The company that developed that technology ultimately became Docker.

At the beginning of 2013, they realized the public PaaS business was a pretty hard business to be in, but the technology could be really interesting. I came on board to advise them about open source — and then as CEO. Docker itself turned a year old on March 20, and my one-year anniversary was April 10. It’s been a whirlwind.

InfoWorld: Why do you think Docker has taken off in such a spectacular way?

Golub: I think we’re solving a really big problem that many people are experiencing. And it’s being heightened by almost every major trend, whether it’s cloud or scale-out or the need for things to be developed iteratively.

When I got started in the business, applications were long-lived, monolithic, built on a single stack, and deployed to a single server. Today, development is iterative and constant, applications are loosely coupled components built on a multitude of different stacks, and they run on multitudes of servers. Somehow, the same complex application built on different stacks has to run in testing, staging, and production; scale across a cluster; move to a cloud; go to a customer; work on a VM; etc.

InfoWorld: That’s a hard problem.

Golub: Yes, it is. What Docker basically does is let you take any application and its dependencies, put it into a lightweight container, and then run it anywhere — or anywhere there is a Linux server. It’s a simple yet revolutionary idea that I think the world has been waiting for, because the current technology is a real mismatch with how people want to run things.

InfoWorld: How do you envision this new container disrupting the market?

Golub: I think it disrupts several things. First of all, at a very basic level, you go from a process of going from development through testing and staging and production that today generally takes weeks, with things breaking at every stage and fingers being pointed, to something that can now take minutes and work 90 percent of the time. And in the 10 percent of the time it isn’t working, it’s really clear whether it’s a development issue or an ops issue. So it’s revolutionizing how people are building code.

It’s also revolutionizing how things can get deployed, because you’re not trying to treat an application as if it were a server and moving around something big and bulky and hard to change. Instead, you’re deploying lightweight containers that can be deployed in milliseconds anywhere — and destroyed just as easily, or updated just as easily, which is a revolution in deployment.

I think we’re also revolutionizing how applications get managed. Because the container is so lightweight and contains its application and can get built directly from source, you know exactly what’s running where, what version it is, and you can update it. So you solve three really big problems while at the same time separating what rightfully belongs with application management from infrastructure management.

InfoWorld: Where did the inspiration for the technology come from?

Golub: Container technology is not new. Almost every PaaS out there was using some kind of container technology, but containers were hard to use and they weren’t portable between different environments. Solomon Hykes, who founded the company, had this insight: Wow, if we actually make this available to developers and we make it easy to migrate between different environments, this can really revolutionize the world.

It was clear that we would do this as open source. Then we could get it integrated into all the new stacks and have something that can run not only on any Linux server, but also easily integrate with DevOps tools, work inside OpenStack, as well as be adopted by other PaaS vendors and by cloud guys. And lo and behold, that’s kind of what’s happened.

InfoWorld: How is your container technology different?

Golub: The analogy we like to use is the shipping container. It used to be that anything you ever tried to ship was in some specialized container: Coffee beans were in bags and car parts were in crates, and you had to unload and reload every time you sent from a ship to a train to a truck to a crane. Things would interact badly, like if you were shipping bananas next to animals.

The shipping container revolutionized all that by being a standard size and shape and having hooks and holes in all the same places. Suddenly, anything can go inside it, you seal it up, and the same box goes from the ship to the train to the truck to the crane without being changed. World trade was revolutionized because, suddenly, all these things were the same. The manufacturer doesn’t really care whether it’s going to go on a boat or a train. He doesn’t even have to know in advance because it’s inside the container.

So that’s kind of what we’ve done. Basically, you just put the application in the container and run it directly on the host. There’s no guest OS. It’s very lightweight. It’s the same thing that you do with an Android phone and its applications, only now it works in the back office as well.

InfoWorld: To what degree do you feel this disrupts platform lock-in?

Golub: We make it really easy to … create things and move them around. So people who built a business model based on keeping you locked in through some artificial means will have problems. If you are an infrastructure provider and you provide the best infrastructure with great security and great uptime at a reasonable price, the fact that it’s easy for people to move stuff to you using Docker should be a good thing. We lower the walls and people who have the best garden will attract people. Trying to have a walled garden in this era isn’t going to work.

InfoWorld: You’re not currently a container for every ship.

Golub: We are a container that will work on any Linux server. It doesn’t matter whether it’s Red Hat or Ubuntu. It doesn’t matter whether it’s physical or virtual. It doesn’t matter if it’s Amazon or SoftLayer or Rackspace. It doesn’t matter if it’s staging, testing, or production. So we’re on a lot of ships and trains and trucks and cranes.

InfoWorld: All the Linux distros are falling in line?

Golub: Yeah. Actually, we could work on them before. Now we’re getting baked in with them. We’ll be shipped with RHEL and shipped with Ubuntu and shipped with Debian; we’re shipped with Amazon Linux AMI and we’re in OpenStack. That’s just a matter of making it easier. Before, people had to download a Docker host.

InfoWorld: Will you stop with Linux?

Golub: There is no fundamental reason why we have to stay in Linux. We can also manage BSD Jails or Solaris Zones, which are sort of the equivalent low-level technology for Solaris, and we have some stuff in the works for .Net as well.

InfoWorld: Really? For .Net?

Golub: It won’t be this year, but …

InfoWorld: That’s going to be interesting. A lot of people have been saying there’s no equivalent technology for Windows, and they don’t think there needs to be one because the software management situation is very different.

Golub: Yeah. People don’t tend to really get it. The other thing that I think is important is that it’s not just technology. It’s also the ecosystem. So, for example, if you go to Docker today you’ll find that there’s an index with over 9,000 applications that were created in Docker. You’ll find solutions, some official and some not official, to work with Chef and Salt and Puppet and Ansible and Jenkins and Travis.

InfoWorld: How else do you plan to extend the functionality of Docker?

Golub: We’re building solutions that make it easy to link containers together, to migrate containers between different hosts, and to see what’s running where. So that’s a lot of our long-term business model, sort of providing the vCenter equivalent for Docker. In essence, as an open-source company, we’ve given away ESX. And the additional value will be in the orchestration and management layer.

InfoWorld: Is there a timeframe for when some of these tools will be available?

Golub: A lot of the very basic things that enable you to link containers together are already out there. The tools that make it possible to orchestrate between different containers in a data center are already there. We are going to be announcing a lot of things at DockerCon on June 9 and 10.

InfoWorld: Has the success of Docker exceeded your expectations?

Golub: It has completely exceeded our expectations. We thought it would take several years to catch on, that developers would like it but it would take a while for sysadmins to embrace it as well, and even longer for more conservative organizations to adopt it. And what has just astounded us is that everything in our multiyear plan has kind of moved up.

We knew developers would love us, and we’re just thrilled that they’ve loved us this much and have piled on this quickly. We’re thrilled by the fact that we now have 400 contributors to the project. Our company is 30 people and a turtle. So having that is spectacular. We’ve been amazed that sysadmins have embraced this almost as enthusiastically as developers. We’re in production at lots of places even though we’re not yet at Docker 1.0.

InfoWorld: Docker 1.0 is next month, isn’t it?

Golub: Either next month or probably at DockerCon.

InfoWorld: What, for you, is the big thing that will signify 1.0?

Golub: There were a few really big things that we wanted to achieve before we got to Docker 1.0. One is that with every release we’ve been shrinking what’s in the core and building a pluggable framework around it. So changes that we want to make or new functionality we want to add, like networking and storage and things like that, can be delivered as plug-ins rather than requiring people to upgrade. At 1.0, we will be at a place where the core doesn’t need to change that rapidly.

Secondly, quality and bake time and documentation. People have been using Docker in production since 0.5. But when a more conservative company adopts it, we don’t want them to struggle with documentation or with rough edges. And finally, we want to be able to offer commercial support. So when we announce Docker 1.0, we’ll be confident that it’s a version that can be supported for the long term.

InfoWorld: So in terms of the business, you would say you’re in “investment mode” now?

Golub: Certainly there are more dollars leaving every month than are coming in. We do sell some tee shirts, and we actually just launched hosted private registries. So what I will tell you is that the growth rate has been phenomenal — but starting from zero, growth is pretty easy.

But the business model is fairly straightforward. What we do is very similar to the Red Hat model: providing commercial support. But we also think that there is a natural set of managed services around orchestration, management, and monitoring that makes sense this year. And at some point in 2015 we know enterprises that want those things delivered on premise, and we’ll do that as well.

InfoWorld: So manage the services through the cloud?

Golub: Yeah, right. Through a hosted service that we provide to make it easy to publish, to find, to download, to sign and create, to move things between different clouds, to move things that have been on-premise to the cloud. And there are actually a lot of services like New Relic and others that have sort of established this model. Enterprises are willing to have certain management functions provided in the cloud, provided that the data itself is still resident on-premise.

InfoWorld: Do you have a sense of the types of applications people are using Docker for right now?

Golub: There’re four major areas of interest. The biggest one, I’ll say, is sort of CI/CD [continuous integration/continuous delivery], generally speaking. So people who want to be able to go really quickly from development all the way through to production. So people like eBay and others, and RelateIQ and others, talk publicly about what Docker has done to revolutionize that.

The second major use case is people who are looking at some kind of a hybrid cloud deployment, where they’re looking for an easy way to either develop in private and move to the cloud or develop in the cloud and move private (or burst between). And Docker provides a really great framework for doing that. And now the major cloud providers are also supporting Docker.

The third major use case we’re seeing is what’s called big data scale-out, where a VM was never appropriate. If you’re trying to do computation across hundreds of machines and scale out and then scale back just as quickly, something really lightweight that’s easy to create, easy to throw away, is the right model.

The fourth is people who are offering multitenant services and are using Docker as a way to do that. So this is like Baidu, which is China’s Google.

InfoWorld: Website hosting?

Golub: It’s actually there for their PaaS. So they’re creating a PaaS based on Docker. Yandex did the same thing — the Google of Russia. And now the Google of Mountain View is sort of doing the same type of things.

InfoWorld: So development on the public cloud is obviously a big part of this momentum.

Golub: Yeah. But I think in general … a developer tends to start with a personal project, stateless, loves it, and then brings it into the organization for simple apps. And then very quickly a sysadmin sees it and says: Oh, we could actually use this in production as well. And then they start thinking about more complex apps where each of the components is containerized and linked together. That’s the general trend that we’re seeing.

InfoWorld: A foundation for distributed applications?

Golub: Yeah. Containers not only provide the right level of abstraction from the application in the host, but they also provide the right level of abstraction between containers. So that if you want to have a complex multitier app, you can put the application and the database and the data each in separate containers, and move them around as appropriate. We’ve provided the tools that let you orchestrate or define how containers interact.

InfoWorld: What open source license do you use?

Golub: We’re Apache. We went as open as we could. Apache is the most permissive license; it’s open design.

InfoWorld: You already have quite an ecosystem out there of people who are doing stuff.

Golub: There are some 350 projects built on top of Docker and at least 20 startups that we know of that are sort of Docker-based. We’re actually one of the largest projects on GitHub.

InfoWorld: That’s a claim to fame in itself.

Golub: It truly is a community driving this. Given that 95 percent of the contributors to the project don’t work for Docker, Inc., we have to be pretty humble about who’s really driving this project forward.

InfoWorld: What sort of app deployments is Docker not good for?

Golub: If you’ve written an app that has specific requirements on specific features in a specific version of the kernel, Docker is not going to be that helpful to you. It’s certainly not designed to let you run a Mac program on a Windows box or vice versa. You want to use a VM for that. If you’re wedded to the notion of state, you want to be using a VM.

InfoWorld: As CEO of Docker, what keeps you up at night?

Golub: Work. I could say I sleep like a baby — which means I’m up screaming. The pace is going so rapidly, and I think the expectations that have been placed on us are so high that I really just want to make sure that we do a great job delivering against it.

This article, “Docker CEO: Our container goes anywhere,” originally appeared at InfoWorld.com. Read more of Eric Knorr’s Modernizing IT blog. And for the latest business technology news, follow InfoWorld on Twitter.

Eric Knorr

Eric Knorr is a freelance writer, editor, and content strategist. Previously he was the Editor in Chief of Foundry’s enterprise websites: CIO, Computerworld, CSO, InfoWorld, and Network World. A technology journalist since the start of the PC era, he has developed content to serve the needs of IT professionals since the turn of the 21st century. He is the former Editor of PC World magazine, the creator of the best-selling The PC Bible, a founding editor of CNET, and the author of hundreds of articles to inform and support IT leaders and those who build, evaluate, and sustain technology for business. Eric has received Neal, ASBPE, and Computer Press Awards for journalistic excellence. He graduated from the University of Wisconsin, Madison with a BA in English.

More from this author