Distributed Computing Studies Get Google, IBM Support
- By Will Kraft
- October 16, 2007
Google and IBM are providing resources to universities focusing on Internet-scale application studies. The resources will help bolster the universities' large-scale distributed computing curricula plans and foster knowledge among computer science students. The companies will provide hardware, software and services to help universities develop and expand research in this area.
The University of Washington was the first institution to join the project, closely followed by Carnegie-Mellon University, MIT, Stanford, UC Berkley and the University of Maryland. The project may expand to include other academic institutions.
Changes in computer architecture are the driving force for why Google and IBM are supporting these educational initiatives. Increases in network speed and capacity permit developers to address problem solving in new ways.
For instance, in order for Web applications such as search, social networking and mobile commerce to work quickly, it's necessary to distribute the processes across many different machines simultaneously, according to an announcement issued by the companies. Parallel programming is useful for resource-intensive tasks, and developers can use that technique for various applications, such as apps for gene sequencing, cryptography and climate modeling.
Google and IBM created a Beowulf cluster, currently made up of several hundred machines (nodes), which creates a work environment for the students. Eventually, the cluster will be expanded to more than 1,600 active nodes. Each node will be equipped with a toolkit of open source software, such as Linux, Xen virtualization tools and Apache's Hadoop Project (an open source implementation of Google's published computing infrastructure). Students will test their parallel computing projects by accessing the cluster over the Internet.
The cluster is already being put to work despite its unfinished state. At the University of Washington, computer science students have managed to produce a program that uses the processing power of the cluster to scan Wikipedia article edits for spam. Another program organizes global news articles by geographic location.
It's an area poised for future growth, and students will get the preparation for it.
"We're aiming to train tomorrow's programmers to write software that can support a tidal wave of global Web growth and trillions of secure transactions every day," stated IBM's CEO Samuel J. Palmisano.