MapR Targets Persistent Storage For Containers
Application containers get a lot of attention, but remain ephemeral. Converged data platform specialist MapR Technologies is looking to fill that gap in the application container ecosystem with the roll out of a persistent storage capability designed to move stateless containers to a full state status to access database tables and files.
The availability of the storage capability addresses a pressing need for production deployment of containers: persistent storage has emerged as a barrier to wider adoption of containers in production.
MapR, San Jose, Calif., announced general availability on Tuesday (Feb. 7) of its converged data platform for Docker containers. The company noted that the struggle to move container applications to production has so far yielded stateless web applications. Lacking data persistence, MapR and others assert, container applications cannot scale. They also lack the resilience to survive routine infrastructure failures.
Until now, file and block storage were among the few options for deploying stateful but less secure container applications.
The company also argues analytics apps and other micro-services must be stateful, that is, aware of historical information and context as well as current status.
The MapR persistent storage capability is itself a container dubbed the Persistent Application Client Container. PACC includes a pre-built Docker container providing application access to files, database tables and message streams. Meanwhile, secure authentication also is provided at the container level. PACC is available now on Docker Hub as a Dockerfile.
Company executives stressed in an interview that PACC would support both legacy and emerging micro-services. "It's all part of the same general trend: better use of data in the enterprise," said Jack Norris, MapR's senior vice president for data and applications.
Norris also noted that the persistent storage architecture remains the same on-premise or in the cloud, allowing users to run existing applications in containers. It also means stateful containers can be restarted or moved across datacenter nodes, for example, to access data. Multiple applications and analytics can share the platform's data repository while computing and storage resources are scaled for each workload.
MapR also was quick to note that its persistent storage approach is not a container orchestration tool. "We are a standard Dockerfile," Norris noted. Instead, the storage platform works with leading container orchestrators such as Docker Swarm, Kubernetes and Mesos. The combination is intended to provide reliable persistent storage, the company claims, addressing a key requirement for production workloads.
Among the other challenges, according to Dale Kim, MapR's senior director of industry solutions, are creating stateful capabilities for running existing applications like databases. Stateful applications are becoming increasingly critical to "event-driven micro-services," Kim added, and emerging container platforms are making it easier to deploy micro-services.
"This product emerged really from customer demand," Norris added. "It's already been pressure-tested among existing customers."
Other approaches exist for providing persistent storage for application containers, but MapR argues they are narrower and "don't lend themselves to supporting [both] existing and new applications," Norris said.
The MapR storage platform along with other recent advances such as standard "plumbing," or runtimes, and other container management tools are filling out the evolving application container ecosystem as prototypes move to production to deliver data-driven distributed applications. Norris said that container management remains a "huge gap" in the Docker infrastructure, a gap the company claims to plug.