MVAPICH Job Startup Performance

• Experimental Testbed (TACC Ranger System): Each node of the testbed is a SunBlade x6420 blade with four Quad-core AMD Opterons (@2.00GHz) and 32 GB main memory. An InfiniBand full-CLOS fat tree connects all the nodes.
• Running time of `MPI Hello World' program on 1K cores is around 4 sec and on 32K cores is around 80 sec.

(Courtesy TACC)