Robert,
epoll won't help at all because in order to use it I need to issue
syscall that have a big latency when the system is loaded (e.g. when
capturing many packets). The problem is not the method (epoll could be
more efficient than poll) but the latency involved.
Cheers, Luca
Robert wrote:
> Dear Dr. Deri,
>
> Thanks a lot for your great paper "Improving Passive Packet Capture" -
> a lot of useful
> information.
>
> My feeling is that it makes sense to use instead of poll() for the
> input demultiplexing
> /dev/epoll for better performance on Linux.
>
> Sincerely,
> Robert Iakobashvili
> Go Networks
>
>
>
--
Luca Deri <deri@ntop.org> http://luca.ntop.org/
Hacker: someone who loves to program and enjoys being
clever about it - Richard Stallman
epoll won't help at all because in order to use it I need to issue
syscall that have a big latency when the system is loaded (e.g. when
capturing many packets). The problem is not the method (epoll could be
more efficient than poll) but the latency involved.
Cheers, Luca
Robert wrote:
> Dear Dr. Deri,
>
> Thanks a lot for your great paper "Improving Passive Packet Capture" -
> a lot of useful
> information.
>
> My feeling is that it makes sense to use instead of poll() for the
> input demultiplexing
> /dev/epoll for better performance on Linux.
>
> Sincerely,
> Robert Iakobashvili
> Go Networks
>
>
>
--
Luca Deri <deri@ntop.org> http://luca.ntop.org/
Hacker: someone who loves to program and enjoys being
clever about it - Richard Stallman