Taking note of how customers have been working with its Hadoop distribution, Cloudera has expanded the scope of its software so that it can serve as a hub for all of an organization's data, not just data undergoing Hadoop MapReduce analysis.
Some of Cloudera's enterprise customers have "started to use our platform in a new way, as the center of their data centers," said Mike Olson, Cloudera's chairman and chief strategy officer.
"We think this is a very big deal. It will change the way the industry thinks about data," Olson said.
Cloudera has released a new beta of its commercial distribution, Cloudera Enterprise, that provides tools for managing an organization's data, as well as tools from Cloudera and third parties for data analysis.
Olson announced the beta of Cloudera Enterprise 5 at the O'Reilly Strata-Hadoop World conference, being held this week in New York.
"It used to be that an organization had lots of balkanized data silos," Olson said. "The stuff that you used to run on a data warehouse because you had no choice, now you can run on the hub."
Putting the data in a Hadoop-based storage repository has many advantages, Olson argued. You can run different types of analytical workloads against the data in the hub. It can easily feed data to other systems, such as content management systems. It can work as an archiving system.
An enterprise data hub, Olson said, can store data as it is generated, even if the organization isn't sure how the data will be needed. Such data may be valuable later for machine learning analysis or other uses not considered.
An enterprise hub also puts security and governance mechanisms in place to safeguard the data. Cloudera has been working on these tools for several releases, Olson said.
"Our ambition is to draw more workloads in and make the hub more valuable over time," he said.
Part of Hadoop's newfound ability to act as a data hub comes from software additions in the latest version of the open-source software, Apache Hadoop 2, on which Cloudera Enterprise is built.
The inclusion of YARN (Yet Another Resource Manager), for instance, allows Hadoop to handle multiple analysis applications, not just those that run on the batch process-oriented MapReduce.
To facilitate the hub, Cloudera has also set up a management framework that third-party analysis applications can plug into. SAS, Revolution Analytics, Syncsort and other organizations have ported some of their software to the platform. Porting analysis software requires that the operations be executed in parallel, as data in Hadoop is typically distributed across multiple nodes, Olson said.
Cloudera Enterprise 5 also adds the ability to cache HDFS (Hadoop Distributed File System) contents in the working memory of a server, which can boost query response and data processing times.
The company's Navigator auditor tool now allows analysts and data modelers to search, explore, define and tag datasets. Users can add customized queries to Cloudera's Impala SQL engine. And Cloudera Enterprise 5 can work with the NFS (Network File System) nodes, which should make the process of injecting data into HDFS much easier, Olson said.
The software also now can take snapshots of the data, providing a backup if the original data is lost or destroyed.