No question about it: Docker's app container system has made its mark and become a staple in many IT environments. With its accelerating adoption, it's bound to stick around for a good long time.
But there's no end to the debate about what Docker's best for, where it falls short, or how to most sensibly move it forward without alienating its existing users or damaging its utility. Here, we've turned to a few of the folks who have made Docker their business to get their takes on Docker's good, bad, and ugly sides.
One hardly expects Steve Francia, chief of operations of the Docker open source project, to speak of Docker in anything less than glowing terms. When asked by email about Docker's best attributes, he didn't disappoint: "I think the best thing about Docker is that it enables people, enables developers, enables users to very easily run an application anywhere," he said. "It's almost like the Holy Grail of development in that you can run an application on your desktop, and the exact same application without any changes can run on the server. That's never been done before."
Alexis Richardson of Weaveworks, a virtual networking product, praised Docker for enabling simplicity. "Docker offers immense potential to radically simplify and speed up how software gets built," he replied in an email. "This is why it has delivered record-breaking initial mind share and traction."
Bob Quillin, CEO of StackEngine, which makes Docker management and automation solutions, noted in an email that Docker (the company) has done a fine job of maintaining Docker's (the product) appeal to its audience. "Docker has been best at delivering strong developer support and focused investment in its product," he wrote. "Clearly, they know they have to keep the momentum, and they are doing that by putting intense effort into product functionality." He also mentioned that Docker's commitment to open source has accelerated adoption by "[allowing] people to build around their features as they are being built."
Though containerization itself isn't new, as Rob Markovich of IT monitoring-service makers Moogsoft pointed out, Docker's implementation makes it new. "Docker is considered a next-generation virtualization technology given its more modern, lightweight form [of containerization]," he wrote in an email. "[It] brings an opportunity for an order-of-magnitude leap forward for software development teams seeking to deploy code faster."
What's less appealing about Docker boils down to two issues: the complexity of using the product, and the direction of the company behind it.
Samir Ghosh, CEO of enterprise PaaS outfit WaveMaker, gave Docker a thumbs-up for simplifying the complex scripting typically needed for continuous delivery. That said, he added, "That doesn't mean Docker is simple. Implementing Docker is complicated. There are a lot of supporting technologies needed for things like container management, orchestration, app stack packaging, intercontainer networking, data snapshots, and so on."
Ghosh noted the ones who feel the most of that pain are enterprises that want to leverage Docker for continuous delivery, but "it's even more complicated for enterprises that have diverse workloads, various app stacks, heterogenous infrastructures, and limited resources, not to mention unique IT needs for visibility, control and security."
Complexity also becomes an issue in troubleshooting and analysis, and Markovich cited the fact that Docker provides application abstraction as the reason why. "It is nearly impossible to relate problems with application performance running on Docker to the performance of the underlying infrastructure domains," he said in an email. "IT teams are going to need visibility -- a new class of monitoring and analysis tools that can correlate across and relate how everything is working up and down the Docker stack, from the applications down to the private or public infrastructure."
Quillin is most concerned about Docker's direction vis-à-vis its partner community: "Where will Docker make money, and where will their partners? If [Docker] wants to be the next VMware, it will need to take a page out of VMware's playbook in how to build and support a thriving partner ecosystem.
"Additionally, to drive broader adoption, especially in the enterprise, Docker needs to start acting like a market leader by releasing more fully formed capabilities that organizations can count on, versus announcements of features with 'some assembly required,' that don't exist yet, or that require you to 'submit a pull request' to fix it yourself."
Francia pointed to Docker's rapid ascent for creating its own difficulties. "[Docker] caught on so quickly that there's definitely places that we're focused on to add some features that a lot of users are looking forward to."
One such feature, he noted, was having a GUI. "Right now to use Docker," he said, "you have to be comfortable with the command line. There's no visual interface to using Docker. Right now it's all command line-based. And we know if we want to really be as successful as we think we can be, we need to be more approachable and a lot of people when they see a command line, it's a bit intimidating for a lot of users."
In that last respect, Docker recently started to make advances. Last week it bought the startup Kitematic, whose product gave Docker a convenient GUI on Mac OS X (and will eventually do the same for Windows). Another acqui-hire, SocketPlane, is being spun in to work on Docker's networking.
What remains to be seen is whether Docker's proposed solutions to its problems will be adopted, or whether another party -- say, Red Hat -- will provide a more immediately useful solution for enterprise customers who can't wait around for the chips to stop falling.
"Good technology is hard and takes time to build," said Richardson. "The big risk is that expectations spin wildly out of control and customers are disappointed."