MapR Technologies, the start-up that surfaced from stealth mode a few weeks ago as the Hadoop distribution in EMC's Greenplum Big Data widgetry, has just put its code out in both a free and paid version, as promised.

It claims it's "the industry's easiest, fastest and most dependable distribution for Apache Hadoop." And, let's face it Hadoop ain't an easy thing to work with.

Cranky rival Cloudera, which employs Hadoop creator Doug Cutting and says it has tens of thousands of nodes under management, complains that MapR isn't open enough because certain features it's added aren't free and none of its source code - free or paid - is open. MapR in turn claims Cloudera doesn't address customer concerns like single points of failure, snapshots and mirroring, and poor performance.

Ah, yes, there's nothing like a good old-fashioned, pot-shot-taking feud to get the blood going.

MapR is supposed to have two-five times the performance of the five other versions of Hadoop out there including the open source project. It also claims to bring "unprecedented dependability to MapReduce analytics."

It's also supposed to take half the hardware required by other distributions, broadening Hadoop's usually high-end user base.

Ticking off cost reduction, dependability and ease-of-use in addition to performance, Enterprise Strategy Group VP Julie Lockner says MapR's "innovations will allow more people to use the power of big data analytics for a range of applications. Both small departmental applications and larger operational deployments will benefit from these advanced data management features opening up opportunities for significant market growth."

MapR says its M3 Edition, free for an unlimited number of nodes, is a complete distribution that includes core performance, ease-of-use and dependability breakthroughs. It's aimed at users of both small and large environments.

The paid M5 Edition includes high availability, data protection and 24x7 support. The widgetry is an easy upgrade to the EMC version.

MapR Technologies claims a passel of industry firsts like no single points of failure throughout the entire system; direct access NFS, the popular protocol used to get data in and out of the cluster; mirroring and snapshots for built-in backup, de-duplication and point-in-time recovery; a so-called heatmap to graphically provision and administer large numbers of nodes and other resources in both small and large clusters; and lockless storage services specifically designed for Hadoop and Big Data's anticipated future growth that enable full random read/write operations across multiple readers and writers.

MapR, which has been working on the widgetry for two years at a cost of $9 million, says it has listened to a lot of customers, partners and the community "about where Hadoop needed major investments to engineer the innovations and breakthroughs that could make a real contribution to advancing and growing the entire market."

MapR has also kicked off an Advantage Partner Program to help customers maximize Big Data analytics and foster an ecosystem and fuel Hadoop's market adoption as the Big Data analytic platform.

The street price of M5 is reportedly about $4,000, putting it in line with Cloudera. MapR will put the stuff out as software and an appliance. It's reportedly been tested by 15 beta customers.

No changes are required to Hadoop applications. The stuff includes the latest patches.

MapR has previously said it's looking at a $500 million market and can save users $8,000 a node long-term.

Downloads are at www.mapr.com/download.