On Wed, Apr 1, 2015 at 10:39 AM, Reindl Harald <h.reindl@thelounge.net> wrote:


Am 01.04.2015 um 19:36 schrieb Dave Johansen:
On Wed, Apr 1, 2015 at 1:32 AM, Reindl Harald wrote:

    Am 01.04.2015 um 06:53 schrieb Dave Johansen:

        I added the call to mlockall() (it did have to be run as root)
        on a F21
        machine with no swap and the slow down was still visible in the "CPU
        bound task"

    surely, as expected

    http://serverfault.com/questions/12679/can-anyone-explain-precisely-what-iowait-is

I don't get how that is expected. The "CPU bound task" is now basically
guaranteed to not be doing anything that requires disk access, so the
iowait should be 0 for the "CPU bound task"

the purpose of mlockall() was not to solve the problem, it's intention is only to reduce side-effects in the test scenario

Ok, that makes sense. I was just confused about the link to iowait and assumed you were implying that it was expected that the "CPU bound task" be impacted by iowait.