Adam Moody edited this page Oct 14, 2016 · 17 revisions

mpiFileUtils is a suite of MPI-based tools to manage large datasets, which may vary from large directory trees to large files. High-performance computing users often generate large datasets with parallel applications that run with many processes (in some cases millions). However those users are then stuck with single-process tools like cp and rm to manage their datasets. This suite provides MPI-based tools to handle typical jobs like copy, remove, and compare for such datasets, providing speedups of up to 20-30x.

The suite is built around a common library currently called "bayer".


Overview of libbayer library

All routines and types prefixed with "bayer_", constants and macros prefixed "BAYER_".

bayer_util.c/h

  • routines to print errors, debug messages, and abort
  • routines to allocate and free memory (bayer_malloc, bayer_strdup, bayer_free, etc)
    • allocation routines call abort and print file and line number on failure (no need to check return value)
    • valid to allocate 0 bytes, NULL is returned (no need to check size > 0 before allocating)
    • bayer_free routine takes a pointer to a pointer, it sets the pointer to NULL before returning (good for catching memory bugs)
    • valid to pass NULL or pointer to NULL to bayer_free (nothing is done in that case)
    • macros like BAYER_MALLOC automatically attach file and line number

bayer_io.c/h

  • wrappers around POSIX-IO routines to retry in case of common, non-fatal errors (e.g., EINTR)

bayer_path.c/h

  • routines to build and manipulate file paths as components
  • routines to simplify paths, e.g., /a/b/../c/. --> /a/c

bayer_flist.c/h

  • distributed list of file info, currently two types: list of (name, type) tuples or (name, stat) tuples
  • can build a list with "walk", "read", and "subset", cleanup with "free"
  • iterate over elements by index number (index local to process)
  • some global "summary" properties of list, e.g., file_max_name, global_size, global_offset
  • functions to query properties of elements, e.g., get_name, get_type, get_atime, etc.
  • see drm/drm.c and dwalk/dwalk.c for examples on how to use bayer_flist
Clone this wiki locally
You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.