Yahoo! Ratchets! Open! Source! Grid! Platform!
Stuffed elephant rising
Hoping to further the Google-battling ways of an open source cluster computing platform named for a stuffed elephant, Yahoo! has decided to share a 4,000-CPU, 1.5 petabyte grid with some boffins at Carnegie Mellon University.
The open source platform is called Hadoop. Yahoo! is the primary contributor to this Apache Software Foundation project, and in opening its grid to Carnegie Mellon, the web giant is pulling in some clever collaborators.
Some Carnegie Mellon researchers will use the grid to build Hadoop-happy apps, and others will help streamline the platform itself. "Basically, we're letting them go to town on this project," said Jay Kistler, the Yahoo! VP who oversees the company's grid computing efforts.
Hadoop was cooked up by Doug Cutting, famous for developing Lucene, the open source retrieval library. In building a search engine called Nutch - yes, it's open source too - Cutting realized it needed some sort of distributed computing platform running behind it. Luckily, Google had given the world a peak inside its Google File System (GFS) and MapReduce technologies.
"A lot of the design for Hadoop came from some papers published by Google," Cutting told us. "These were very nice descriptions of some very useful sounding systems - though they didn't tell us everything we needed to know, so we did diverge from Google's designs quite a bit."
Mirroring GFS and MapReduce, Hadoop offers a distributed file system as well as a means of pooling processing power. And both are designed for use with clusters of low-cost machines. "You can use commodity hardware, rather than storage appliance-type solutions," Cutting explained. "So standard PC motherboards, standard processors, standard drives."
Cutting dubbed the platform Hadoop after his son's yellow stuffed elephant. "I was looking for a name that wasn't already a web domain and wasn't trademarked, so I tried various words that were in my life but not used by anybody else. Kids are pretty good at making up words."
The platform soon caught Yahoo!'s eye, and Cutting is now a Yahoo! employee. He stresses that Hadoop is still an open source project managed by the Apache Software Foundation, but Yahoo!'s role shouldn't be underestimated. The company is already running Hadoop on some unknown portion of its back-end infrastructure. "It's used as a component - or supporting element - on some of our production systems," Kistler told us.
Meanwhile, the company's research lab has donated a new programming language to the project. Known as Pig, it's a way of building apps for the MadReduce-esque portion of the platform. "Think of it as a SQL for MapReduce," Cutting said. This will play a big role in the work at Carnegie Mellon.
Thanks to Yahoo!, the university now has access to a computing grid spanning 4,000 processors, three terabytes of memory, and 1.5 petabytes of disk space. The company says it ranks as one of the top 50 supercomputers in the world, hitting 27 trillion calculations per second at peak performance.
And if things go well at Carnegie Mellon, Yahoo! will open this grid to other universities as well. ®
Sponsored: Becoming a Pragmatic Security Leader