Title:  A Survey of Fault Tolerance in Message Passing Systems

Abstract: The Message Passing Interface (MPI) is the de-facto
message-passing standard in parallel applications running large-scale
distributed systems. But the standard of MPI does not cover fault
tolerance. Recently, various implementations of MPI have addressed
fault tolerance at several levels with different mechanisms. This
survey analyzes the architecture and the fault tolerance mechanisms of
several MPI implementations, such as FT-MPI, LAM/MPI, MPICH-V and
OPEN-MPI.