"Alfred Perlstein" <[EMAIL PROTECTED]> wrote:
> I think you probably want to experiment with pools attached to the
> pipe, and you ought to be using pipe rather than socketpair.
My tests indicate that pipe performance in this case is identical to
socketpair performance. Perhaps because I'm sending so little data over
the wire (in the real thing, most of the data is in shared memory, the
pipes just provide the queuing).
> Meaning
> have a group of children share a pipe, but not more than N, where N
> is where you start to see performance problems.
Actually, the hit comes rather quickly, 8 procs per pipe causes half the
hit, even 4 is substantial. In any case, once I've gone away from having
one pipe, I've already lost the advantage and might as well have a pipe per
worker. I'd much rather manage one pipe than 32 pipes, but managing 4
pipes versus 32 pipes is less compelling.
Thanks,
scott
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
