Mitigating Depreciation of High-Performance Computers with Opportunistic Computing

The Holland Computing Center (HCC) at the University of Nebraska–Lincoln recently acquired a 21-teraflop supercomputer. Even as local usage grows to fill this large resource, a significant portion of the machine currently remains underutilized. Cluster hardware depreciates on a three-year schedule (a standard warranty period), dwarfing operational costs. HCC has thus implemented the Open Science Grid software stack to allow OSG virtual organizations to utilize otherwise idle resources. Cluster utilization has hovered near 100% on a regular basis since implementation of the OSG stack was completed. The opportunistic grid jobs serve as a "real-world health monitoring" for cluster resources.