With the Internet of Things generating anticipation about a new model of distributed computing, MapR Technologies Inc. is hoping to get a jump on its big-data rivals with a version of its Converged Data Platform built specifically for processing at the network edges.
MapR Edge, which is being announced today at Strata + Hadoop World, is essentially a miniature version of the company’s big data platform that can run on small clusters of at least three nodes with as little as 16 gigabytes of memory and 64 gigabytes of disk storage. The three-node minimum is intended to provide a base level of redundancy and parallel processing. MapR said it will consider supporting smaller devices if customer demand merits it.
“There’s a full database, file system and streaming system that sits at the edge,” said Jack Norris, senior vice president of data and applications. “It’s a small form-factor processing capability that’s part of a broader distributed deployment model. The administration is centralized. The intent is that this is something you set and forget.”
MapR is responding to growing interest in a new computing architecture that distributes intelligence to the network, primarily in the form of intelligent devices. The assumption is that IoT-intensive environments will generate too much data to be processed economically from a central location, so decision-making needs to happen closer to the intelligent device.
Research firm Wikibon, which is a sister company of SiliconANGLE, has called edge processing “a vital component of IoT.” Venture capitalist Peter Levine has recently suggested that a distributed computing model could eventually supplant the cloud.
A post-cloud model?
MapR Edge is being positioned as complementary to the company’s big data platform with special accommodations for security, latency and sporadic network availability. “This is not a separate product that we have a light connectedness to; it’s a full, integrated extension of the core product with security, data protection and other things that are gotchas when you extend to the edge,” Norris said.
The company’s big data platform has all the necessary features to handle replication, synchronization and coordination of data streams from multiple endpoints and so doesn’t need to be modified to accommodate the endpoint nodes, he added. “All of the data complexity you’d expect to encounter we’ve already built in,” Norris said.
Applications will need to be modified or created from scratch for such a distributed environment, and MapR doesn’t intend to get involved in that side of the business. Norris said the company will rely upon its partner ecosystem and existing development toolkits to adapt to IoT-intensive environments. Containers and microservices are ideal for deployment in such settings, he said. The company intends to offer on-demand training for the new platform.
Pricing hasn’t been set, but Norris implied that it will be aggressive. “Our intent is that you’ll have many of these small edges, so we expect to bring it in at a much lower price point,” he said. “It’ll be priced to be part of a package with the core.” The product is available immediately.