Containers, Shmontainers! VM Forks (What?) will Rule the Enterprise

Sorry, Docker.

Between Canonical’s efforts to build a container daemon for its operating system and the recent introduction of a homegrown application packaging format from CoreOS, it would seem that the race to bring the new way of virtualization into the enterprise is only now heating up. But the open-source community is competing over a prize that has already been taken.

VMware decided the outcome of the derby three years and eight months ago, when it awarded a computer scientist from the University of Toronto with a grant to research the applications of forking in virtual environments. Three years before that, associate professor Eyal de Lara helped produce a paper describing a mechanism for cloning images across multiple hosts with a small fraction of the overhead normally involved in the process.

Instead of copying the entire contents of an instance as many times as the total number needed, the virtual machine is used as a template for “children” that start out including only the bare minimum necessary to begin execution and share common files with the parent from there on. That greatly reduces hardware requirements, making it possible to squeeze considerably more clones on any given server while reducing instantiation from minutes to potentially less than a second.

If that sounds similar to containers, that’s because it is. Although instances created under the paradigm are technically called “VM forks,” the main functional benefits are identical to what Canonical, CoreOS and Docker offer with their respective projects. The difference is that the technology runs at the hypervisor layer instead of directly on the host, which provides the incomparably better security and manageability of conventional virtualization but without the cost of shuffling duplicate operating system images around.

VMware has been working to implement the mechanism in its software as part of Project Fargo, a low-key initiative that made its debut in August to promises of faster provisioning for virtual desktop environments. That’s just the tip of the iceberg, however. The ability to rapidly spin up a cluster without waiting minutes for each node to come online and generating massive amounts of redundant data in the process makes it much easier to scale distributed workloads, which opens up an entire universe of use cases beyond taking the pain out of the morning boot storm.

That would make for a highly compelling pitch even if VMware didn’t already have roughly half of the world’s virtual servers under its thumb. But since it does, the company’s argument for using forks over containers will also include compatibility with existing infrastructure and integrated management; music to the ears of budget-strapped CIOs. Then in parentheses are the practical advantages to buying from an established supplier instead of trying to cobble together something similar out of open-source components that have only been around a few months.

In other words, containers have little chance of loosening VMware’s grip on the enterprise. But Microsoft might. The Redmond giant won a patent for “applying the concepts of forking and migration to virtual machines” almost a year before Eyal de Lara and his colleagues published their paper, which spells competition in the virtualization giant’s future — as well as potentially tie-breaking litigation.

Regardless of how that almost inevitable lawsuit shapes out, the incumbents will make it immensely difficult for the likes of Docker to establish a foothold in on-premise production environments. Where the movement has an opportunity is the public cloud, which VMware and Microsoft can expect a harder time reaching with their hypervisor-specific forking capabilities. The balance of power in the everyday data center, however, will remain slanted firmly towards the traditional virtualization camp through the container era.

—

Banner via Best Business Intelligence