The practice of I/O optimizations for out-of-core computation

Out-of-core computation needs high I/O performance as the out-of-core data stored in files has to be moved between memory and disks during the execution of an application program. In order to obtain better performance of out-of-core programs, two I/O optimizations are practiced, which are data prefetching and data reuse. This paper focuses on discussing the mechanism and implementation of each method and under which condition a method should be used. Several sets of experiments are taken to test the strategies and the results show that the I/O performance is efficiently improved by using these optimization methods.