Scale-out storage vendor Coho Data Inc. today said an upcoming software upgrade will enable customers to run persistent, portable storage containers directly in primary storage.
Coho Data said a DataStream software upgrade later this year will allow customers to schedule, launch and execute Docker containers directly on its DataStream hybrid and all-flash storage nodes. Running containers at the storage layer would let customers run analytic databases and other services atop existing data, according to the vendor.
The next generation of DataStream software will include application interfaces that enable Docker containers to interact directly with underlying storage, said Andrew Warfield, CTO at Coho Data. The vendor will orchestrate network connectivity between Docker instances using the Google Kubernetes framework and software-defined network switches in DataStream arrays.
Warfield said containers sidestep performance and management issues associated with virtual machine storage.
"A lot of people working on containers have been fixated on compute. The thing that I've always wanted us to pursue with Coho is to build an enterprise storage system that would enable you to do a whole bunch more active things. We're allowing people to add new services on top of their existing data. They will be able to push code into the storage system that lets them analyze, transform and present the data they already have in new ways," Warfield said.
Shared namespace makes DataStream containers persistent, portable
Open source Docker containers package Linux application libraries with a shared operating system. They are commonly used in application development, but concerns about portability and data persistence have slowed adoption in in enterprise storage.
Coho Data container images will be hosted in a Docker registry in DataStream for instantiation by tenants within the data storage system. The system will fire up requests for containers, as new objects are added to storage.
DataStream Docker containers provide persistent storage by mounting a subset of the storage system namespace, which also enables containers to move from node to node with application data intact. An automated scheduling tool in DataStream lets a container stop on one node and restart on another, retaining access to its data on the originating node.
Randy Kerns, Evaluator Group
"We manage the placement of containers internally in the system. We do it partially in response to available CPU, and partially in response to where the data [being used by the container] lives," Warfield said.
"Anytime you execute directly on the node, where your storage software functionality resides, you have to be careful not to compromise the integrity of the data," Kerns said. "The key element here is the registry of containers. What Coho has done is protected the data by using a container to isolate and be selective about which data-in-place apps they run on the node."
Docker leads the pack for containerization in the cloud
Primary storage vendors gradually add Docker container support
Docker container storage and virtual machine storage differ