With the surging popularity of containers creating a host of new companion platforms and accompanying management headaches, a startup founded by a team of former Citrix Systems, Inc., Microsoft, and VMware Inc. engineers and backed in part by former VMware CTO Steve Herrod is coming out with an orchestration platform that it hope will make large-scale container deployments fashionable in the enterprise.
Billed as “vSphere for Containers,” ContainerX is intended to do for containers what dynamic resource scheduling (DRS) did for virtual machines nearly a decade ago: Make them scalable, manageable and elastic. The technology enables developers to allocate resources directly from the Docker command line to create shared clusters of containers in a multi-tenant environment that enables resources to scale dynamically and minimizes the potential for corruption and over-contention.
The problem ContainerX is trying to solve originates with the immaturity of containers and the profusion of orchestration, networking and persistent storage products that have sprouted up to manage them. A typical multi-container environment lacks the multi-tenant features that keep individual container instances out of each other’s way. Resource allocation is limited, meaning that contention for things like storage, memory and CPU can slow performance or even crash containers. Containers also don’t scale out very well because resources allocated to them are fixed at launch.
ContainerX says it’s solved all these problems with a concept it’s calling dynamic resource scheduling. Working initially on a bare metal server, the platform enables the user to create a single cluster that can be shared between pools of containers based upon how much of the aggregate resource each pool needs. For example, one could pool can be assigned 30 percent of CPU and 20 percent of memory, and another assigned 40 percent of CPU and 50 percent of memory. ContainerX says the beauty of its approach is that those resources can be scaled dynamically depending on workload, and aggregate allegations can even be assigned to exceed 100 percent on the assumption that not all containers will run at full throttle all the time.
“If one pool has a spike in workload, the other two pools back down if they’re assigned a lower priority,” said Kiran Kamity, a former RingCube Technologies, Inc. executive who heads up the new venture. “The system is automatically elastic. You don’t have to worry about taking from one pool to give to another. “
ContainerX is also rolling out a management platform that Kamity said provides “a single pane of glass for all your containers.” The company recommends running its platform inside an operating system hosted on a bare metal server and specifying where to deploy containers across Windows, Linux, cloud and physical destinations. All container instances can then be managed centrally. The initial version of the software will support bare-metal, Amazon Web Services and Linux hosts. Windows support is set for the first quarter of next year.
The company’s basic stack includes Docker for containers, Docker Swarm for orchestration, Socketplane for networking and a proprietary persistent storage module it built from scratch. “We’re totally open to leveraging other parts of the ecosystem, but for version 1.0 we had to make some choices,” Kamity said. “If our customers strongly request using a different technology, we’ll be open to that. “
The company’s success is by no means assured. It faces competition from VMware, Red Hat Inc. and Amazon, among others, all of which are adding container supporting their existing management suites. But with $2.7 million in seed capital and an experienced management team that includes Pradeep Padala, one of the architects of DRS on VMware, it has credibility without the encumbrance of a legacy installed base.
Pricing has not been set, but Kamity said it will be an annual subscription model, probably based upon the number of concurrent containers. A free option will be available for environments with less than 50 containers.