Thanks All,<br><br>Only recently I could have a time to change the configuration of switch hub and test this and that. Unfortunately, I failed to take the advantage of Open-MX, which could not run Cactus benchmark for some reasons.<br>
<br>But after changing the maximum frame size to 9000 and enabling flow-control, the latency could be reduced down greatly,<br><br>## mpi_pong reuslts<br>...<br>...<br><br> 262144 bytes took 708 usec ( 740.412 MB/sec)<br>
524288 bytes took 1500 usec ( 699.101 MB/sec)<br> 1048576 bytes took 3523 usec ( 595.256 MB/sec)<br><br> Max rate = 1139.390288 MB/sec Min latency = 2.980232 usec<br><br>##<br><br>And also I could get reasonable scalability for Cactus benchmarks. The MTU of the ethernet cards should be taken 1500, the default one. If I increase the value, it produces worse results.<br>
<br>In conclusion, the culprit of the scalability problem was the interconnect I used, not the cpus or softwares. I appreciate all your comments and suggestions.<br><br>Regards,<br><br>Hee Il<br><br>