Have you tried usiing POE::Loop::Epoll?

David

On 7/5/06, Michael Giambalvo <[EMAIL PROTECTED]> wrote:

I'm working on a large scale TCP server application.

Based on profiling, I see that more than half the CPU time of my
server session is spent either in POE::Wheel::ReadWrite->read or
POE::Wheel::ReadWrite->write.

Any suggestions on eliminating this bottleneck?  My session is
currently implemented using POE::Component::Client::TCP.   I guess I
could implement my session in C, but I'm not sure how I'd do it in a
way that avoids the same overhead the POE::Wheel::ReadWrite suffers
from.

Anyone have experience with implemented high volume clients with POE?

Reply via email to