The Ohio Supercomputer Center recently awarded computing systems to nine Ohio university faculty members as part of the Cluster Ohio Projectóan initiative led by OSC, the Ohio Board of Regents, and the OSC Statewide Users Group to encourage the state''s university faculty to build local computing clusters.
OSC is upgrading its own cluster to SGI Intel Itanium systems, and launched Cluster Ohio to distribute the processors from the original configuration.
Thirty-four research proposals were received in the competition for academic cluster computers. The nine winners included three bioinformatics-related projects: Charlotte Elster and Peter Jung of Ohio University for "Modeling Complex Systems on a Beowulf Cluster," John Gallagher and colleagues from Wright State University for "Cluster Computing for Bioinformatics and Biocontrol," and Austin Melton''s group from Kent State University for "Scientific Visualization and Modeling."
Grant recipients will receive two to four systems of two or four 550 MHz Intel Pentium III Xeon processors. The clusters are scheduled to be delivered in September and October. In addition to the processing units, OSC will provide on-site maintenance, software, training, and systems administration advice to grant recipients.
Elster said the 24-node cluster at Ohio University would support three separate research projects. Since a single researcher would only be eligible for 16 processors, "we joined forces so we all get more out of it," Elster said. Jung said his work on simulating ion channel clusters for neurons and heart cells is ideally suited to cluster computing.
"One of the reasons we proposed this for Cluster Ohio is because these simulations are very good for parallel code," Jung said. Because the ion channels essentially share only one variable, he anticipates a "tremendous" linear speedup. "I expect if we have 24 nodes it will be 24 times as fast," said Jung.
Jung said he is looking forward to using the cluster for the next stage of his research, simulating whole brain tissue cultures to determine how glial cells communicate with neurons.
At Kent State, Melton''s cluster grant will be used to support a number of faculty modeling projects, ranging from describing the processes and functions of a prokaryotic cell to predicting predator-prey interactions in lake Erie.
"The complexity of such models is often too great for most computing systems," said Melton. "Since clustered machines are as computationally efficient as very large and expensive computers for many problems, this award will allow us to make substantive advances in our modeling efforts."
At Wright State, Gallagher and his colleagues are also splitting cluster time between a number of projects. Gallagher, Travis Doom, and Michael Raymer of the university''s bioinformatics group combined their proposal with another project in order to secure access to a 16-processor cluster.
Gallagher''s group singled out three projects that would see the greatest benefit from parallel computing. The first will parallelize an evolutionary computation search for promising drug lead compounds based on protein binding information; the second will apply parallel clustering and regression algorithms to the analysis of multiple microarray data sets; and the third involves building an accurate heart model to improve the control of artificial cardiac pacemakers.
In addition to their local systems, grant recipients will also have access to the 160-processor Itanium cluster the OSC installed in May. An additional 160 processors are expected to be added to the OSC cluster in the fall.
OSC plans to upgrade its cluster every 24 months and expects to issue a similar call for proposals through the Cluster Ohio program every two years.