Yahoo! exposes very own stuffed elephant code
Distributed data-crunching distro
Hadoop Summit Yahoo! has released its own Hadoop distro, an internet-scale distributed data-crunching platform based on the Apache open-source project that underpins several of the web’s highest profile sites, including Yahoo!, Facebook, and - amusingly - Microsoft’s Bing.
Inspired by Google-published research papers describing Mountain View’s proprietary software infrastructure, Hadoop is the brainchild of open-source guru Doug Cutting, the Nutch crawler founder who’s now on the Yahoo! payroll.
Yahoo! has used Hadoop code on its production infrastructure for more than a year now, and after calls from the ever-growing Hadoop community, the company is opening up its internal implementation of the project.
"We’ve put a lot of investment on our testing and deployment," Yahooligan Eric Baldeschwieler said Wednesday at the Yahoo!-sponsored Hadoop Summit in Santa Clara, California. "We’re going to take that work that we put into it and put it out on the web."
The new release - known as the Yahoo! Distribution of Hadoop - is not a commercially supported distro. "We’re not getting into a new business," Baldeschwieler explained. Yahoo! is leaving that business to Cloudera, the Silicon Valley startup that unveiled a commercial Hadoop distro this spring.
According to Baldeschwieler, Yahoo! will release code identical to that tested and deployed on the company’s internal machines. "The source code release will be exactly like we use on Yahoo! clusters," he said. And he expects Yahoo!-tweaked code will be released three to six months after general release of the Apache project code it's based on.
Yahoo! will not restrict access to the code, which will be available here from the Yahoo! developer network. It will merely require an agreement before downloading. The first release will be Hadoop version 0.20, which is now under alpha test inside the company.
Yahoo! contributes about 72 per cent of all Apache Hadoop patches. And it now uses Hadoop code to crunch data for myriad Yahoo! services, including its search index and the automated system that chooses news stories for its homepage.
Cutting cooked up Hadoop in 2004, naming the project after his son’s yellow stuffed elephant. Along with Yahoo!, one of its early users was Powerset, the semantic search engine recently acquired by Microsoft. Powerset now drives at least a small portion of Redmond’s latest Google challenge, which it insists on calling Bing. ®