Anticipating greater Apache Hadoop use in production environments, Cloudera has outfitted its commercial distribution of the data processing framework with additional configuration and management tools.
"As more and more Hadoop use cases are moving to production, people are starting to care about service level agreements and the quality of service provided to end-users," said Charles Zedlewski, Cloudera vice president of products.
[ Discover what's new in business applications with InfoWorld's Technology: Applications newsletter. ]
Version 3.5 of Cloudera Enterprise updates the management suite with two modules, each of which helps manage Hadoop deployments.
One new module, called the Activity Monitor, provides "a real-time view into the performance of all the running workloads in a Hadoop system," Zedlewski said.
Performance monitoring is typically found with standard relational databases, though tools for Hadoop remain scarce, he noted. This module displays the status of various Hadoop components, including MapReduce jobs, Hive queries, Pig scripts and Oozie workflows. It tallies about 45 metrics overall, including CPU, memory and network usage, as well as statistics on scheduling.
Cloudera also introduced a service and configuration manager module. This software "provides a single means of managing the Hadoop stack," Zedlewski said. The Service and Configuration Manager automates many of the processes for updating these programs and provides a set of checks that can be used to verify that the changes requested did indeed take place.
A typical Hadoop deployment running across 30 servers may be comprised of about 95 programs, with a configuration file to go along with each program. As a result, any administrative change might require restarting any number of these programs, or making changes to multiple configuration files.
Using this software, "It's much less likely that you will incur downtime due to mis-configuration or mis-setting," Zedlewski said.
With this release, Cloudera also updated a number of existing modules as well. Users can now track historical usage of disk and file space usage within the Resource Manager. Users can also now administer (ACLs) Access Control Lists in the Authorization Manager module.
Although most Hadoop tools are open source, Cloudera's management modules are only available as part of Cloudera's paid edition of Hadoop. However, the company is making available for no cost a version of the Service and Configuration Module, called SCM Express, that will help potential users set up and test a Hadoop deployment.
This version of the module can download all the appropriate programs and set them up on designated servers. "You can be running your first Hadoop workload from a Web client within 10 minutes," Zedlewski said.
Large ISPs such as Yahoo, Facebook, and eBay, use Hadoop for analyzing user behavior, and the venture capital investment community anticipates its wider use in the enterprise. Cloudera Enterprise is a pre-integrated stack of Hadoop software and associated support available on a subscription basis.
"You take a look at how you operationalize any platform. People typically want to get some visibility into what is going on," Zedlewski said. "This is the first time people have been able to apply that full lifecycle management to a Hadoop system."