Parallel I/O scheduling in multiprogrammed cluster computing systems

In this paper, we address the problem of effective management of parallel I/O in multiprogrammed cluster computing systems by using appropriate I/O scheduling strategies. The performance of I/O devices lags behind the performance of processors and network systems, resulting in I/O becoming the bottleneck in current systems. This gap is expected to increase in the future since I/O performance is limited by physical motion. Therefore, it is imperative that novel techniques for improving I/O performance be developed. Parallel I/O is a promising approach to alleviating this bottleneck. However, very little work exists with respect to scheduling parallel I/O operations explicitly. We propose new I/O scheduling algorithms and evaluate the relative performance of the proposed policies against the most commonly used approaches. Our preliminary results show that the proposed policies outperform current methods and can substantially enhance the performance of real-world scientific applications.