Parallel Compilation doesn't work between nodes

Hi All,
       I compiled LAMMPS by intel mpi and mkl. The compilation works for parallel calculation using cores of one node but doesn't work for cores on different nodes. The following is the error which is followed by the makefile for my compilation. Any ideas about what has gone wrong? Thanks a lot.

ERROR MESSAGE:

cu63:52b5:d37aeba0: 2299 us(2299 us): ucm_create_services: ERR Cannot allocate memory
cu63:52b4:ba67bba0: 3311 us(3311 us): ucm_create_services: ERR Cannot allocate memory
cu63:52bc:ef006ba0: 4757 us(4757 us): ucm_create_services: ERR Cannot allocate memory
APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)

Hi All,
       I compiled LAMMPS by intel mpi and mkl. The compilation works for parallel calculation using cores of one node but doesn't work for cores on different nodes. The following is the error which is followed by the makefile for my compilation. Any ideas about what has gone wrong? Thanks a lot.

ERROR MESSAGE:

cu63:52b5:d37aeba0: 2299 us(2299 us): ucm_create_services: ERR Cannot allocate memory
cu63:52b4:ba67bba0: 3311 us(3311 us): ucm_create_services: ERR Cannot allocate memory
cu63:52bc:ef006ba0: 4757 us(4757 us): ucm_create_services: ERR Cannot allocate memory
APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)

these are low-level messages, not LAMMPS messages. most likely they
originate from the MPI library. this indicates either a configuration
problem, a permissions problem or a hardware problem. you need to
contact the people managing the cluster that you are running on about
this.

axel.