Apache project paves path to easier Hadoop

Ambari 2.0 promises a lot, but competition remains strong from the proprietary offerings in the field

Apache project paves path to easier Hadoop

If there's one point people can agree on when it comes to Hadoop, it's the difficulty in setting it up and keeping it running. An Apache project dedicated to such work, Ambari, has only gone so far.

The 2.0 release of Ambari, however, promises more of the features Hadoop admins have been itching for -- assuming they haven't already elected to go with a proprietary solution to Hadoop management.

The "what's new" section of Ambari's official project site provides only a bullet-point list of the changes in Ambari 2.0. But a few catch the eye, among them automated rolling upgrades, and a new alerting framework and metrics system (the last of which replaces the Ganglia project).

Assuming those changes deliver what's promised, these features ought to appeal to Hadoop admins who have struggled with such functions in the past.

Most of this new work on Ambari ostensibly stems from a collaboration between two of the bigger vendors of Hadoop distributions, Pivotal and Hortonworks. Pivotal had its own proprietary Pivotal Command Center product, but has since elected to collaborate with Hortonworks on expanding support for Ambari, with Pivotal providing support for Ambari as an alternative to its own product.

The bigger question is whether customers of Hadoop distributions will opt for Ambari, especially if they have alternatives.

Cloudera, arguably the largest of the commercial Hadoop vendors, offers its own management tools. (Cloudera's no slouch as a contributor to Hadoop, but nothing says it has to contribute to all of Hadoop.) Those management tools may appeal more to enterprise users who want things to simply work.

Ambari may also be one of the factors in Cloudera not joining the unified-Hadoop Open Data Platform project. Cloudera had its own reasons to be skeptical of that initiative, but according to another argument advanced by Silicon Angle, there was another possible angle: "Cloudera’s challenge is to stay ahead of the market by providing superior functionality to the open source equivalents of products like Cloudera Manager," says that article, and joining the OPD would have meant adopting Ambari and "essentially ... capitulating to Hortonworks."

If Hortonworks and its other partners can keep Ambari competitive, it'll provide a reason to choose a Hadoop distribution that doesn't swap out parts of the Apache family of products with proprietary replacements. But the size of Cloudera's existing user base and the long-standing roughness of both Ambari and Hadoop both stand in the way.


Copyright © 2015 IDG Communications, Inc.