Original URL: http://www.theregister.co.uk/2011/06/29/mapr_announces_two_hadoop_offerings/

MapR unleashes two 'next-generation' Hadoop distros

Well, one next-generation, and one free semi-next

By Cade Metz

Posted in The Channel, 29th June 2011 19:31 GMT

MapR Technologies – a Silicon Valley startup that spent the last two years revamping Hadoop for use in the enterprise – has unveiled two new ditributions of the distributed number-crunching platform.

On Wednesday, amidst Yahoo!'s annual Hadoop Summit in Santa Clara, California, MapR announced a free offering known as the Map3 M3 Edition and a for-pay version known as MapR M5. Both are based on the open source Apache Hadoop project, but MapR has rebuilt certain portions of the platform in an effort make Hadoop both more nimble and more reliable. The company likes to say it has delivered the "next generation of Hadoop".

The free M3 distro can be used on an unlimited number of nodes, but it lacks the certain Hadoop enhancements available with the M5 edition, which can be licensed for $4,000 per node per year. These are the first Hadoop products directly available from MapR. Previously, the company announced that its technology will underpin the upcoming Hadoop offerings from EMC.

Though MapR refers to its two new products as "distributions", these are not entirely open source. MapR has made extensive changes to the underlying platform that are proprietary. That said, the company recently signed a Corporate Contributor License Agreement with the Apache Foundation, signaling its intention to contribute code back to the community, and vice president of marketing Jack Norris tells The Register that such work is already underway. "It's an evolving process," he says.

Unlike Apache Hadoop, both MapR distros can be mounted on a standard network file system (NFS). "What we've done is rearchitect the storage services so that they provide random read and random write for multiple readers and writers, and then we expose that as an NFS mount, so in addition to being able to use that data from Hadoop APIs, you can use all your standard Linux tools and Unix tools and applications," MapR CEO and cofounder John Schroeder has told us. "You can create real-time data streaming out of Hadoop. You can make Hadoop look like a big C: drive on your Windows desktop."

Both distros also include the MapR "Heatmap", a graphical interface for managing Hadoop nodes. The company claims that both products provide much higher performance than the open source incarnation of Hadoop. And in addition offering the core Hadoop File System (HDFS) and Hadoop MapReduce platforms, both products include sister platforms such a HBase (a distributed database), Hive (a query language), and Zookeeper (a means of managing distributed services). There are all existing open source platforms, but according to Norris, MapR has "pre-tested" all these Hadoop pieces in concert.

Though the free M3 distro can be used on an unlimited number of nodes, it does not include the new Hadoop job tracker and name space built by MapR. These are only available with the for-pay M5 edition.

Typically, the Hadoop Job Tracker – which distributes jobs across a cluster and manages their execution – is a single process running on a single machine. With the M5 edition, MapR has introduced a Job Tracker that's distributed across several machines. With Apache Hadoop, the global namespace is also a single point of failure, and the MadR distributes this NameNode as well.

The MapR M5 edition also includes what the company calls 24-by-7 support. ®