Follow Datanami:
December 7, 2011

MapR Update Extends Hadoop Accessibility

Datanami Staff

This week MapR released an updated version of its Hadoop distribution, adding some features that are targeted at improving overall performance and accessibility for a larger set of applications.

The release of Version 1.2 is focused on extending user access, both in terms of their diverse applications and environments, not to mention reaching out to those who are still waffling about which distribution is the best fit.

In addition to expanding API access, MapR says they are including support for MapReduce 2.0 for when it becomes available. While it’s still some time off before the newest release of Hadoop is production-ready, the company says “users will be able to take advantage of the combined benefits of MapReduce 2.0, such as backward-compatibility and scalability and MapR’s unique capabilities, such as HA (no lost tasks or jobs during a failure) and the high performance shuffle.”

MapR is working to address some of the stability issues that have plagued some Hadoop users, issues that have lent to some deciding to keep the open source version out of production environments. In the new update they have upgraded a number of elements, including Hive, Pig and HBase. They also claim that they have been able to identify several “critical stability data corruption issues” in HBase that have been fixed.

For those who are still on the Hadoop distribution fence, the company announced that it is now possible to access an entire MapR cluster as a free VM to experiment with the platform and “try before buying” into the Hadoop distro. Using the test virtual cluster will allow potential users to play with some of the elements that make MapR a bit different, including its NFS capabilities and snapshots. They claim that testing out the distro can happen within minutes on a standard laptop.

The company plans on rolling out the new version later this week and will make the test cluster available at the same time. Many of the companies pitching Hadoop distributions are looking for ways to let users make the differentiations themselves, it seems that opening access and making onboarding simple is one of the only ways of accomplishing this.