Dear All,

    I've built a cluster to test the IO500 benchmark.

    The preliminary results showed that the new "shift" parameter did influence the stat result badly.

    If I use "-N" parameter, the mdtest result is:

        mdtest_easy_stat    1429.930 kiops : time 253.86 seconds

        mdtest_hard_stat      962.500 kiops : time 191.10 seconds

    If I don't use "-N" parameter, the mdtest result is:

        mdtest_easy_stat    200250.000 kiops : time 1.84 seconds

        mdtest_hard_stat      51038.200 kiops : time 4.47 seconds

    It is clear that the "shift" parameter change will impact the stat result with the same hardware and software configuration. Now I'm still trying to optimize the test. Hope I'll deliver a better result later.

Ruibo Wang