And on the sixth day (of June) MapR announced its new streaming distribution would use Spark and not MapReduce – though this will complement rather than replace its Hadoop distro.
Although San Jose-based MapR tips its hat to MapReduce by name, the increasing obsolescence of Google's 2004 framework – and the public enthusiasm for Apache Spark as its successor – has provoked the company into developing its own “enterprise-grade Apache Spark Distribution.”
This will include “the complete Spark stack” the company says, alongside its own IP in what it terms the MapR Converged Data Platform, to offer customers speedy in-memory processing, speedier app development, and code reuse across those applications.
MapR is also going to include its Spark Distribution in its plug-and-play "Quick Start Solution" Hadoop offerings, which first came out last year to flog pre-built templates, configuration, and installation help.
In a canned statement, Anoop Dawar, product management veep at MapR, said: “We’ve seen significant growth of customers deploying Spark as their primary compute engine. We believe this gives our customers a converged compute and storage engine for batch, analytics, and real-time processing that helps build and deploy applications rapidly.”
Corresponding with El Reg, Jack Norris, MapR's new senior veep for data and applications, said: "There is a lot of excitement in the developer community around Spark."
MapR is seeing more growth in its free on-demand training classes, which relate mainly to Spark, and Norris added: "Developers talk about the ease of development in Spark and [say] the streaming analytics options are very strong."
Norris said that a "hybrid open source model that can combine architectural innovations while supporting industry standard APIs and supporting the full rich open source community is the best model for meeting customers’ needs." ®