Data management based on MPI and high performance IO
Zhao Hanwen · 2016
Because of the huge amount of data and computation, the field of seismic exploration is a hot research topic in the field of high performance computing. With the advent of the processing mechanism of continuous growth and multiprocessor systems, the handling capacity of the computer system itself obtained the rapid development, but the growth rate of the I/O performance cannot keep up with the development of system processing capacity. Thus, I/O becomes one of the main bottlenecks of high performance parallel computing. To solve this problem, the basic idea of I/O in parallel computing environment is an important breakthrough point. Even in the parallel computing environment, the basic method of I/O system also includes serial I/O and parallel I/O two classes according to the different access technology. I use the MPI library technology this paper implements parallel I/O, and provides the corresponding parallel data management strategy. According to the location of the data in the file is continuous or not, we use two methods to achieve parallel I/O, in order to improve the parallel environment for the rapid positioning of the data and fast reading and writing.