Brian Bockelman |
Over the past two years, Brian Bockelman has qualified as a Ph.D. candidate in both mathematics and computer science at the University of Nebraska-Lincoln, set up and maintained a working grid computing site for UNL physicists collaborating on a particle physics experiment, helped researchers at three other universities get up to speed on the grid, and—most recently—celebrated his twenty-first birthday.
"I started out as a mathematics graduate student," says Bockelman, "but after a year and half wondered, since I was spending so much time with computers, why not go ahead and do both?" One year ago, as the first degree candidate in a new joint mathematics and computer science program, Bockelman started working on the UNL grid site. The university had been chosen as a "Tier-2" computing site for the CMS particle physics experiment, which meant that a cluster needed to be set up on the Open Science Grid. Bockelman installed the middleware and software necessary for computation, storage and data transfer.
"At the beginning it was just me and four new servers," Bockelman recalls. "A lot of the grid stuff is still very fragile, so at first it took quite a bit of effort to keep everything running, but there's been a marked improvement over the past year." In addition to improvements to the stability of grid middleware, the past year has also brought two additional full-time staff members to the UNL CMS site.
In addition to graduate classes in computer science and maintaining the grid site, Bockelman has been developing extensions to the MonALISA monitoring system in the hopes of creating a package that other CMS grid sites will use in the future. His work with other universities has been part of an effort to create a regional virtual organization with several other Great Plains Network members. Together with other computer scientists from UNL, Bockelman helps researchers at the Universities of Arkansas, Kansas and Missouri learn about grid computing and the Open Science Grid in preparation for launching the new virtual organization by the end of the summer.
By then, Bockelman will have ramped up work on his dissertation research. He plans to create a strong implementation of sinc methods, a family of numerical techniques based on the sinc function. These numerical methods give accurate approximations of derivatives, definite and indefinite integrals and convolutions.
"Sinc methods have a lot of really nice theoretical properties and will work on a wide variety of problems," explains Bockelman, "but you can't go anywhere and download a sinc method library; it's not packaged or freely available anywhere. The functions are a little intricate to apply and difficult to use, so we want to come up with something general that people can plug in and use. We hope the implementation will be something that will scale to the grid."
—Katie Yurkewicz
e-mail this article
|
