From Documentation
Jump to: navigation, search

Doing large-scale computation can be much easier than saving the results of the computation. Given a computation – or a series of computations – utilizing hundreds or thousands of processors, how do we write our code in such a way that we can efficiently read and write our data, and have our results in a more structured format than in hundreds of separate files? This introductory course will present a variety of strategies for doing large-scale Input/Output (I/O) with either parallel jobs or large numbers of serial jobs. In particular we will discuss such Parallel I/O technologies as MPI-IO (part of MPI-2 and up), and parallel I/O libraries HDF5, NetCDF, and ADIOS.