Yahoo! has opened up its Google-battling Hadoop research cluster to three more big-name US universities.
Yesterday, the company announced that The University of California at Berkeley, Cornell University, and the University of Massachusetts at Amherst are poised for large-scale data-crunching research on its M45 cluster, the 4,000-processor, 1.5-petabyte-disk data center opened up to Carnegie Mellon University in the fall of 2007.
The M45 runs Hadoop, the open-source distributed computing platform based on the Google File System (GFS) and MapReduce technologies developed at Google. In 2004, the Mountain-View Chocolate Factory published a pair of papers on these groundbreaking technologies, and they were soon seized upon by a man named Doug Cutting, known previously for developing Lucene, the open-source retrieval library. He called his shadow-project Hadoop, after his son's stuffed elephant.
In 2006, a couple of years after launching Hadoop under an Apache license, Cutting was hired by Yahoo!, and the Google rival is now the project's primary contributor.
On one level, Yahoo! uses the M45 to promote Hadoop research, to foster the ongoing development of the platform. But it's also a means of educating future Yahoo! employees. Like Google, Yahoo! has found that new hires are ill-prepared for the sort of uber-data-crunching that goes on inside its walls.
"We're thinking about long-term recruiting for the company," Ron Brachman, vice president and head of Yahoo! academic relations, told The Reg. "We want to make sure students are educated in modern technology and modern ways of building and running systems, particularly when it comes to very, very large-scale internet-style applications that companies like Yahoo! deal with everyday.
"Not that long ago, these kinds of enterprises were not really in the mainstream computer science curriculum."
Google launched its own Hadoop research cluster in the fall 2007. Mountain View likes the educational benefits of Hadoop. But it also likes keeping MapReduce technology hidden behind its usual code of secrecy - and one step ahead of the competition.
Last year, Yahoo! put Hadoop behind its new Yahoo! Search Webmap, a mega-app that builds a database of all known web pages – complete with all the metadata needed to understand them. According to Yahoo! grid guru Eric Baldeschwieler, the app draws its web map 33 per cent faster than the company's previous system.
And now, Brachman told us that the platform powers the real-time automated algorithms that now select news stories for its home page. During an artificial intelligence conference last summer, Yahoo! claimed these algorithms boast news story click through rates by 25 to 30, making the company millions of extra dollars in ad revenue each year.
Based on the Kalman algorithm – a filtering method developed in the early 1960s – the system determines home page story placement by analyzing many millions of user clicks as they happen.
Famously, Hadoop also underpins Facebook's infrastructure. And Amazon is now offering the platform as a web service over its AWS virtual data center. ®