Current Practice and a Direction Forward in Checkpoint/Restart Implementations for Fault Tolerance