Containers are here to stay. But this raises genuinely tough questions: Where are they staying? How many are there, and in what form? What's the real extent to which containers are becoming a tool of business IT?
A year ago, New Relic published the first report about container use in the enterprise, using data gleaned from its monitoring tools. New Relic has run the numbers once more, and while it hasn't been able to answer some of the big questions, the new findings are still eye-opening.
The more the merrier
In a phone call, Abner Germanow, senior director of strategic marketing at New Relic, discussed one obvious finding: the explosive growth in the number of containers in use.
Over the last year, the average number of containers increased by 192 percent in the companies profiled, from 96 to 280. In 2015, the maximum number of containers seen running in any one organization was 1,596; in 2016, that jumped to 135,630. Germanow also stressed this wasn't a case of one firm skewing the figures; the figures were up unilaterally.
Another key finding involved container lifetimes. Originally, New Relic broke out roughly three lifespans for containers: "pets," or containers that ran for thousands of hours on end; "cattle," containers with lifetimes in the tens to hundreds of hours; and "bacteria" or "mayflies," whose lifespans are measured in minutes and sometimes seconds.
While the categories still stand, New Relic noted that longer-running containers are running longer -- up to 13 days from the previous average of 11 days -- and shorter-running containers are running shorter. Many more containers run short than long, so the average lifespan of all containers has come down, from 13 hours to 9 hours, 15 minutes.
To VM or not to VM?
One common question about container use is whether it's displacing the prevalence of VMs and, if so, under what circumstances.
Germanow didn't feel the data collected by New Relic could support an answer: "It's a really hard thing for us to track," he said. Anecdotally, he noted that New Relic was seeing many instances of containers used as part of the build process, but not as part of production -- another explanation for more short-running containers. That said, he was reluctant to draw larger conclusions based on those facts.
Fintan Ryan, industry analyst at RedMonk, believed short-running containers are part of a shifting pattern of workloads, though not specifically at the expense of VMs. "While not displacing VMs yet," he wrote in an email, "there is definitely a pattern of 'lift and shift' occurring for certain classes of applications, something that we saw with both VMware and AWS in the past."
Ryan also noted the growth in containers seen by New Relic was consistent with RedMonk's observations: "Most, if not all, forward-looking enterprises are looking toward containers as part of their next-generation application strategy."
Another anecdotal revelation from Germanow: New Relic saw containers used as an end run around the restrictions of operations departments. This development is less an indictment of devops as a whole than it is a response to the inadequacies of certain implementations of devops.
"[Dev groups] can't obtain virtual machines or servers as fast as they would like," said Germanow. "There are multiple instances of ops teams getting really excited about improving their pipeline, and watching their requests for new servers or VMs all of a sudden flatline, and when they tried to figure out what happened, they found out the dev teams deployed Docker." Germanow was again uncertain about where those containers were running, but believed they were running on VMs.
As containers have taken off, various firms (New Relic among them) have developed tooling for obtaining better metrics about their use and the behaviors of applications running inside them. But there's been little front-line information regarding the environments around containers or how containers are changing those environments. Such details are hard to come by, but they would be eye-opening to have.