Microsoft, Red Hat, IBM and others help Google build cloud Trojan Horse
Tech companies team up to build Kubernetes container manager
Tech companies large and small are teaming up to develop open-source software Kubernetes, the success of which will reduce the relevance of VMware's virtualization tech in the cloud.
IBM, Red Hat, Microsoft, Docker, Mesosphere, CoreOS and SaltStack announced on Thursday that they are working with Google to develop open-source software that lets developers use Linux Containers rather than virtual machines as the basic components of large cloud-based software systems.
This is a big move with subtle yet far-reaching implications for VMware, because a world ruled by Linux Containers is not a world in which VMware can make vast sums of money.
Linux Containers let developers package up applications into individual secure sandboxes each isolated from the rest; these containers are run side by side on the same machine running on the same Linux kernel. The containers share the same underlying operating system.
Virtualization, meanwhile, emulates computer systems as virtual machines, in which developers can put applications and any OS of their choice. Each virtual machine is entirely separate to the others all the way down to the operating system; this means one VM can run Windows, another Linux, on the same machine, for example.
Linux Containers cost less from a computer resource perspective, and are therefore more efficient to run at scale.
"Everything at Google runs in a container," Google engineer Joe Beda explained earlier this year.
There's one fly in the tantalizing ointment of containers, though, which is that if the security of the underlying OS is compromised then every container running on the machine is also in trouble, unlike virtualization which carves off one OS for each application. Such an exploit against containers surfaced last month though was quickly fixed.
Kubernetes is a tool developed by Google to ease management of Docker's Linux Containers in the cloud. Google launched Kubernetes at DockerCon last month. Though the company says it is based on an internal cluster manager called Omega, El Reg has established that the difference in capabilities means it is fairer to say it is inspired by Omega, rather than being a direct derivative of it.
Nonetheless, it represents Google's view on how best to marshal software applications built around containers in the cloud and will therefore, if developed properly, end up competing with other cloud management systems like vSphere and more.
Microsoft will make sure Kubernetes works for containers based on Linux virtual machines on its Azure cloud; Red Hat will bring it to its cloud products; IBM will contribute code to Kubernetes and Docker; CoreOS will add in support for Kubernetes to its distributed operating system; SaltStack will make it easier to run Kubernetes in other places; and Mesosphere is going to add the tech to its eponymous Mesos open source project, the companies said.
The arrival of Kubernetes on Mesos is an intriguing development as Mesos was a tool developed at Berkeley (and now developed by Twitter) meant to closely ape the advanced capabilities of Google's own Borg and Omega systems.
By adding the two together, Mesosphere says in a draft blog post seen by El Reg that "we believe that Kubernetes on Mesos will marshall in a new age of app development. It will give any company an easy way to adopt the Google-style scale-out model of building apps, as well as provide for improved developer workflow and resource optimization."
One company not mentioned in this release is Amazon – and for good reason. Amazon, via its cloud division Amazon Web Services, dominates the cloud world while keeping all of its internal infrastructure secret. Part of the reason why all these companies are working together, though many compete in the business sphere, is that they want to nullify AWS's lead by co-developing some powerful technology that could take it on.
We're not completely sure that's a viable strategy, though – after all, if "everything" at Google runs in a container, then it's likely Amazon has implemented the same thing. It just isn't saying anything. ®