Avatar

LAM/MPI logoWith a little sadness, I note that LAM/MPI was officially retired recently.

LAM/MPI’s hosting provider, Indiana University, made the decision not to renew the lam-mpi.org domain any more.  As of a few weeks ago, LAM/MPI’s web site is no more, and its domain is in the process of expiring.

LAM/MPI was a highly popular implementation of the MPI standard that was initially developed at the Ohio Supercomputing Center, eventually transferred to the University of Notre Dame, and then later finally moved to Indiana University.

The LAM infrastructure initially targeted transputers, which included a sophisticated run-time system for Linux clusters.  The MPI layer was added after the fact — almost as a publicity stunt to gain more users.  The MPI layer soon became the most popular part of the platform, however, and became the focus of all research and development.

LAM/MPI mainly targeted Linux clusters, but also supported a wide variety of other platforms, including OS X, AIX, and HP-UX.  LAM/MPI supported several different network transports, including TCP sockets, a few flavors of shared memory, Myrinet, and InfiniBand.

Here’s some fun stats:

  • Despite active development stopping around mid-2004, LAM/MPI’s last Subversion commit was r7856, on June 9, 2008.
  • There had only been 7 commits that year, and only 11 in the year before that.
  • LAM/MPI’s last official released version was 7.1.4.
  • Although there were a few beta releases of 7.1.5 (mostly bug fixes for specific users), it was never formally released.

I got my introduction to the world of MPI and parallel computing in LAM/MPI.  I started off as a user, but eventually became one of its primary developers.  Most of the other developers were graduate students and research staff at Notre Dame and Indiana University.  In LAM/MPI’s heyday, it facilitated science and research at hundreds of sites around the world, and was included in every major Linux distribution.

Although development stopped over a decade ago, the spirit of LAM/MPI lives on in Open MPI.  Indeed, many of LAM’s ideas and some of its architecture were used to create the fundamental underpinnings of Open MPI.

So long and thanks for all the fish, LAM/MPI!