It is common at research institutions to maintain multiple clusters. These might fulfill different needs and policies, or represent different owners or generations of hardware. Many of these clusters are under utilized while researchers at other departments may require these resources. This may be solved by linking clusters with grid middleware. This thesis describes a distributed high throughput computing framework to link clusters without changing security or execution environments. The framework initially keeps jobs local to the submitter, overflowing if necessary to the campus, and regional grid. The framework is implemented spanning two campuses at the Holland Computing Center. We evaluate the framework for five characteristics of campus grids. This framework is then further expanded to bridge campus grids into a regional grid, and overflow to national cyberinfrastructure.
Supported by the National Science Foundation and the U.S. Department of Energy's Office of Science