June 10, 2009, 1:20 PM — To advance the adoption of the Hadoop technology for processing massive data sets, Yahoo will release a version of the software it has perfected to power many of its key sites and services.
The announcement was on Wednesday at the second annual Hadoop Summit, organized by Yahoo and sponsored by several major vendors that back this technology, including IBM, Sun and Amazon.com.
Yahoo's search engines, ad systems and webmail service use Hadoop, an open-source implementation of the MapReduce programming model for processing large data sets in processor clusters.
At Yahoo, Hadoop runs on more than 25,000 servers to crunch data on tens of billions of Web pages, tens of petabytes of storage and billions of records every day.
By releasing the source code of its Hadoop distribution, Yahoo hopes to boost development of this Apache Software Foundation project.
Yahoo, which employs Hadoop's founder Doug Cutting, contributes back its Hadoop enhancements to the Apache project, but now is releasing its full Hadoop implementation, which is the largest in the world and has been extensively tested and perfected for use in live, production systems. The Yahoo code comes without restrictions on its commercial use.
"There has been a lot of interest expressed that if we made this available, it would be useful to other users," said Shelton Shugar, senior vice president of cloud computing at Yahoo.
Yahoo is increasingly consolidating processing platforms on Hadoop clusters because they can scale better and provide the large-scale data processing the company's sites and services require, he said.
"We're very happy with how Hadoop is working for us. It's a fabulous platform," Shugar said.
Amazon.com recently launched a hosted service designed to simplify for developers the use of Hadoop called Amazon Elastic MapReduce, while companies like Cloudera market commercial Hadoop distributions designed for enterprise deployments.
