Docker Vs. Kubernetes: Rackspace CTO Weighs In
By Lorna Garey, Editor in Chief
At last week’s OpenStack Summit, SDxE caught up with John Engates, Rackspace’s CTO, and Bryan Thompson, general manager for the Rackspace OpenStack Private Cloud, to get their take on the state of enterprise adoption of Docker versus Kubernetes and what they expect in open storage options as OpenStack and containerization take off.
Judging just by the number of mentions at the Summit, Kubernetes is the cool kid on the next-gen infrastructure playground. But there’s more to it.
First, the comparison: Docker is a container platform, while Kubernetes, a Google spin-out, is a container orchestration engine. Swarm is the Docker orchestrator, but Docker containers can be managed via Kubernetes. Mike Kavis, VP and principal architect for Cloud Technology Partners, offers an in-depth explanation here, suggesting that Kubernetes pushed Docker toward more openness — a good thing for enterprise IT.
Still, the buzz award goes to Google.
“Kubernetes has just taken off,” said Engates. “The communities around them, they caught on like wildfire, and folks see it as even more open than the Docker approach.”
He says Docker veered too close to commercialization in an effort to jumpstart use of containers in the enterprise.
“Sometimes open source communities get a little bit scared — is it too commercialized too quickly, are these people going to own it and control it?” said Engates. “I think Google did a good job with Kubernetes to really keep it as an open community.”
As for where data will rest, software-defined storage is clearly the answer. However, there’s jockeying among Swift, Ceph and native Cinder on open, commodity hardware and more traditional arrays from the likes of Dell EMC — which just announced a major revamp of its storage catalog at its annual Dell EMC World event — and HPE, as well as hot flash startups like Kaminario and Pure Storage.
“Depending on workloads and what you're trying to do, I think we're seeing more and more customers leverage multiple different storage tiers and platforms and use availability zones,” said Thompson. “Do I need high-performance ephemeral storage, do I have low-performance stuff I need for test devs, spinning disc? It's a broad array.”
No pun intended, presumably.
Companies also have specialized workloads and sunk investments in hardware and expertise.
“When you think about private cloud in a customer premises environment, they're already used to buying large-scale storage gear, they've got a lot of expertise built up, and sometimes they just want to continue to use that resource, so plugging it into an OpenStack private cloud makes a lot of sense in that case,” adds Engates.
In terms of performance characteristics, especially for relational databases or other demanding or specialized workloads, commodity hardware and Ceph can’t yet match the price/performance of high-end enterprise storage, especially as the cost of solid state drops and quality rises.
“It's now multi-flavor,” Engates says. “Tiering just makes a lot of sense. Put the right workload on the right piece of storage kit, and that's thankfully something that OpenStack accommodates pretty easily. You've got a lot of ways to solve problems.”