Today, Mirantis joined the Cloud Foundry Foundation. But, unlike HP or IBM, we are not doing it to build a Mirantis Cloud Foundry distribution. We strongly believe it is the best strategic move for Mirantis to remain the pure play OpenStack company, and to actively integrate best-in-class outside technologies into OpenStack. We do this because it is the best way to give our customers choice in how they architect their cloud, and this philosophy applies to the PaaS layer. Whether it’s Cloud Foundry, Docker, OpenShift, Kubernetes, or some other PaaS – we don’t care as long as it runs on OpenStack.
We are also keeping a singular focus on OpenStack because we anticipate the PaaS market evolving towards a collision track that may make many promising startups and technologies irrelevant. “What collisions?” one might ask. “Isn’t it all clear at this point: OpenStack is the open source IaaS; Cloud Foundry is the open source PaaS? The battles have been fought and territories have been captured…no?”
As I am writing this, we are witnessing a shakeout in the OpenStack market in action. Nebula, the once-hottest startup of the OpenStack space, backed by VC heavyweights like Kleiner Perkins, announced yesterday that it ceased operations. The OpenStack Foundation was launched 3 years before the Cloud Foundry Foundation. History is bound to repeat itself with a similar shakeout around PaaS as we are seeing in OpenStack today.
It may not be obvious, but the Docker-fueled container craze is much less of a threat to VMs or OpenStack than it is to PaaS vendors. Container-centric vendors (like Docker) have already learned that monetizing containers to optimize data center density won’t win them many customers. There are only so many enterprises that will run homogenous data center infrastructure at massive scale; the story of “it works for Google, it’ll work for the enterprise” doesn’t work. Container-centric vendors will soon realize that the way to monetize containers is by selling a container orchestration layer, which is no different than selling PaaS.
Let’s take a step back and look at the forces battling for PaaS market dominance:
The Cloud Foundry ecosystem
Today the Cloud Foundry ecosystem consists of Pivotal CF, HP Developer Platform, IBM BlueMix and ActiveState Stackato. One way to look at Cloud Foundry is that it’s the leading open source PaaS. “The Docker way” to look at Cloud Foundry is that it’s a feature-rich, enterprise-grade orchestrator for a container standard that only few have heard of.
Docker originally started as a hosted PaaS called dotCloud, directly competitive with Cloud Foundry. The PaaS market was overheated at the time, and the company realized that they would be unable to capture PaaS mindshare. So they pivoted to Docker, instead capturing developer mindshare around their container standard. But since containers are impossible to monetize, Docker is now looking to monetize “container orchestration,” which is no different from PaaS, and will soon have them competing with Cloud Foundry again.
Docker (the company) is not the only one that wants to use the container wave to win the PaaS game. A whole ecosystem of so-called Docker-derived players have jumped in. All of them are destined to eat at the PaaS market pie:
CoreOS. Once a big supporter and partner for Docker, CoreOS owns the control point for container-optimized Linux. They feel it is a significant enough control point to roll their own container standard, Rocket. Projects to build Docker/CoreOS orchestration are already out there. It’s a matter of time before it becomes an ecosystem of its own.
Kubernetes. A Docker orchestration built by Google, the largest user of containers in the world. Enough said. Rumor has it that Google is putting 20 of its engineers that worked on Omega (the miraculous container engine that runs everything at Google) to work on Kubernetes next quarter.
Red Hat OpenShift. An interesting example of putting many of the independent technology control points into a comprehensive solution that customers buy. The Red Hat native PaaS is getting a second life after losing the battle for PaaS mindshare against Cloud Foundry. Red Hat is throwing all it’s got at PaaS, reworking OpenShift to embrace Docker as its container standard, Google’s Kubernetes for orchestration and Red Hat’s Atomic as a container-native OS. The company is actively contributing to both Docker and Kubernetes and is starting to see some success with OpenShift.
A few others worth mentioning: Mesosphere – a container orchestration platform based on Mesos, originally built by Twitter. Currently positioned as microservices infrastructure for big data apps like Kafka, Storm, Hadoop etc. They won’t admit it, but Kubernetes is a direct competitor. RancherOS – a container-optimized OS (like CoreOS) + orchestration. Started by the original CloudStack crew.
Owns PaaS APIs and sys admin mindshare for on-premise PaaS.
Unsexy technology stack. Warden container standard. No container-native operating system.
Owns developer mindshare for container standard.
Far behind Cloud Foundry on container orchestration story.
Sexy technology stack – Docker, Kubernetes, Atomic. Optimized for Red Hat’s stack.
Optimized for Red Hat’s stack. Brand plagued with “failed technology” stamp.
Owns mindshare for container native OS.
Nothing much further up the stack from the operating system yet.
Built by Google.
Not yet GA.
With so much going on in the PaaS market, at Mirantis we would rather stick to OpenStack, which is what we do best. Again, our customers should be able to choose Cloud Foundry, Docker, OpenShift, Kubernetes, or some other PaaS player – our mission is to make sure it works best on OpenStack. Joining the Cloud Foundry Foundation is an important step in that direction.