We are talking about this on irc at the moment. I had never used POE::Loop::Event and assumed it was libevent, but unfortunately it isn't. We'll get a libevent loop written soon.
David On 6/5/07, David Davis <[EMAIL PROTECTED]> wrote:
FYI, POE does have an libevent loop. POE::Loop::Event Cheers, David On 6/5/07, Bob Maccione < [EMAIL PROTECTED]> wrote: > > Event::Lib is a really nice library for doing simple POE-like stuff. > > We use a lot of POE but I'm moving some code to Event::Lib due to the > max connection concerns. Event::Lib uses the libevent library and seems > as stable as POE in our current testing. For our really critical code > we are using the C api and doing all the work in C. Perl would probably > work but C is just safer from a performance standpoint. > > Note that this is in the Linux/Solaris world, I'm not doing a win32 > stuff. > > -----Original Message----- > From: Rocco Caputo [mailto:[EMAIL PROTECTED] > Sent: Tuesday, June 05, 2007 8:48 AM > To: POE List > Subject: Re: Maximum concurrent TCP connections > > On Jun 4, 2007, at 15:09, Zack Payton wrote: > > > > My first question is: Is 10,000 simultaneous a realistic goal? > > Maybe if you combine it with IO::Poll or Event. The default select() > loop scales badly because of the way select() works. Also, > POE::Component::Server::TCP doesn't use the fastest techniques. It's a > generic module first, and optimized for speed second. > > > Secondly, is this a POE issue or an operating system issue? > > It's not a POE issue. POE doesn't care if you try to exceed Perl or OS > limitations. It'll just report the error or die trying. > > > Is the error mentioned above have anything to do with system level > > file handles or is it, as the source code documentation mentions, an > > attempt to mimic the unix mentality of "everything is a file" and it's > > > actually a windows socket issue? > > No idea there. > > -- > Rocco Caputo - [EMAIL PROTECTED] > > >
