Like many universities, Purdue needs a great deal of
computing tools to conduct research. For instance, these tools have helped render
fly-through animation of a proposed satellite city to serve as a refuge for
Istanbul, Turkey, in case of an earthquake. They’ve also been used to create Nano
Factor, a game designed for middle school-aged students interested in science
and engineering.
Innovations like these mean that computing power is always
in great demand, even if budgets for servers and other power essentials fall
short.
Despite this need, many machines throughout the Purdue
campuses have only been in use half of the time. So university officials have
collaborated with Wethersfield, Conn.-based Cycle Computing to build DiaGrid, a
grid of idle campus computers/servers to provide computing capacity. DiaGrid
serves the main Purdue campus, along with Indiana University, Notre Dame,
Indiana State, Purdue’s Calumet and North Central campuses, as well as Indiana
University-Purdue University Fort Wayne.
With a motto of “No Cycle Left Behind,” those working on the
project selected the free, open-source distributed computing system, Condor,
developed by the University of Wisconsin,
and what’s called the CycleServer compute management tool from Cycle Computing.
Computers in the pool run client software, and efficiently and securely connect
them to front-end servers. Jobs are then submitted to these servers and
parceled out to various pool machines when idle. This way, tens of thousands of
processors can be brought to tackle problems from various researchers. Using
Condor’s flexible policy features, technical staff can control when and how
their machines are used, such as "evenings only."
Results are already surpassing expectations: The university
has increased total capacity to more than 177 teraflops—equal to a $3 million
supercomputer requiring several thousand square feet of data center space. But
DiaGrid came at a cost of just $100,000. With approximately 31,000 processors,
it currently offers 15.6 million available computation hours, and over the last
month, the flagship clusters were 97.3 percent utilized. DiaGrid was recently
named by IDG’s InfoWorld as a “top 100 IT
project” of the year.
The project allows university researchers to continue to
pursue computationally intensive tasks, such as simulating the Oort Cloud to
better understand the solar system’s formation, or project the reliability of Indiana’s
electricity supply. “DiaGrid is an essential component to our research
infrastructure,” says John Campbell, associate vice president for Information Technology
at Purdue. “During the past year, it provided over 17 million hours of
computation for researchers in a variety disciplines, from agriculture to
engineering.”
The project demonstrates that the most expensive servers are
those that are already owned and in place, says Jason Stowe, founder/CEO
of Cycle Computing. “We’ve worked with companies in multiple industries—including
financial services and life sciences—to harvest unused resources, and see this
becoming a more popular cost- and energy-saving trend in the future.”
