On Wed, Mar 9, 2011 at 8:57 AM, Philip Martin <philip.mar...@wandisco.com> wrote: > You are not looking at the right epoll_wait, look 14 lines down from the > start. That epoll_wait call takes <0.039952>. The line below that > shows a delay of 0.040042 to the subsequent read call. So epoll_wait is > responsible for nearly all the delay.
This epoll_wait() is waiting for the response back from the server, so it makes sense that it'd take longer. Do we know the server responded sooner? Could it be something on the server side? IOW, this pattern looks correct on the syscall() side. -- justin
