There’s a lot of theory when it comes to gauging actual adoption of Docker containers. But when it comes to discerning theory from reality, nothing is in a better position than a widely used service for monitoring IT environments.
According to the folks at Datadog 10.7 percent of the 100,000 IT deployments it monitors on behalf of customers have now deployed Docker containers. That may not seem high, but it does represent a 30 percent growth rate year over year.
K. Young, director of strategic initiatives at Datadog, says the most surprising thing about those results is that larger enterprises are leading the way. About a quarter of organizations with 500 or more hosts have adopted Docker. Because these organizations are running at much higher levels of scale, interest in the agility and enhanced utilization rates enabled by Docker is more acute, he says.
Datadog, which just added an Automated Service Discovery capability for monitoring Docker containers, is also reporting that Docker containers are now running on 10 percent of all the hosts it monitors. That’s up from 2 percent 18 months ago. On average, those hosts are each running five containers, Datadog says. In fact, Datadog reports that two-thirds of the companies that try Docker wind up deploying it in some fashion. Once those organizations adopt Docker containers, usage of Docker containers increases by a factor of five within nine months, Datadog also discovered. However, a Docker container only has an average lifespan of 2.5 days compared with 15 days for the average virtual machine, reports Datadog.
The most widely used technologies used in association with Docker are Registry, NGINX web servers, Redis key/value data stores, Elasticsearch engines, MySQL databases, Logspout log management software, the Quay registry software from CoreOS, the etcd key/value store, Postgres databases and CAdvisor software for collecting metrics about containers in Kubernetes environments. Datadog also reports there has been a notable decline in use of the Ubuntu distribution of Linux from Canonical and the MongoDB document database.
Perhaps the biggest takeaway from the list, Young says, is that it’s clear deployment of stateful applications on databases is on the rise. That suggests usage of containers in production applications is becoming more widespread.
Of course, the single biggest impact containers may have on IT is finally forcing the DevOps issue. Interest in more modern approaches to DevOps has been running high, but implementation of best DevOps practices remains limited. Adoption of containers, however, finally may force that issue. Not only do containers tend to turn over faster than virtual machines, the sheer number of them requires IT operations teams to re-evaluate their existing processes just to keep pace. In addition, over time those containers are likely to wind up running on top of bare-metal servers, virtual machines and in platform-as-a-service (PaaS) environments deployed in both public and private clouds.
Unfortunately for many of those IT operations teams, however, far too many of them still tend to find out where those containers are actually deployed long after the fact.