High-Density Protein Microarrays

65297 Large-scale computing, or high performance computing, is moving increasingly towards clusters of commodity computers. As the size of these clusters grows, the number of components involved statistically reduces the mean-time-between-failure of the overall system. Because electronic components do not tend to fail suddenly, the integrity of the components can be checked periodically to identify those that are experiencing errors before they are assigned work. This would minimize the compute cycles wasted by component failure and greatly increase mean-time-between-failure of the overall system. This project will develop runtime hardware diagnostics to identify faulty hardware while a system is idle, thereby preventing failure while running real applications. Phase I will develop a modular diagnostic testing framework and implement prototype hardware independent plug-in modules for testing specific hardware after the system is booted. The modular framework will consist of an architecture neutral hardware test dispatching mechanism. Commercial Applications And Other Benefits as described by awardee: As increasing numbers of clusters, and clusters of increasing size and complexity, are deployed to address the computational needs of the nation, emphasis is beginning to shift from how to assemble these clusters to how to maximize their benefit. Being able to diagnose and map out failing elements before they impact computations should minimize the number of cycles lost to hardware failures and allow users to make greater progress in solving problems.