Clayton O'Neill


Event::Lib::UDPPump - Event::Lib event type for high performance UDP applications.


  use Event::Lib::UDPPump;
  use IO::Socket::INET;

  my $numchildren = 10;

  my $s = IO::Socket::INET->new(Proto => 'udp', 
                                LocalPort => 5000);

  fork() foreach (1..($numchildren - 1));

  my $pump = udppump_new($s, \&callback, @args);

  sub callback {
    my ($results_href, @args) = @_;
    # Process results here.


This module is intended for people writing high performance UDP applications. It is an extension of the Event::Lib module, and can provide better performance in several circumstances.

When a UDPPump event is registered for a UDP socket, a pthreads thread is created in the background. All this thread does is block in recvfrom waiting for a packet to be received on the socket. When that happens, it will then pass a packet over to the main thread which will call the callback you have registered. This will be more efficent specifically in the case when you have several processes or threads all processing data on the same UDP port. This avoids the problem with multiple processes all blocking in select (or poll, etc) waiting for traffic on the same socket, and then all waiting up and trying to read from the socket at the same time when new data arrives.

This can make it easier to implement daemons where each request may require significant processing. This is because while your callback is running, the recvfrom thread will be blocking waiting for you to complete. This means that you can run a number of child processes as workers, and as long as you have a few of them waiting in recvfrom then response times will not suffer.

The other primary benefit is that it can allow you to take advantage of multi-processor/multi-core servers without having to resort to using threads. This is the primary reason that this module was implemented.


udppump_new($socket, $callback, [@args])

$socket should be a UDP socket.

$callback should be a code reference. When a packet is received it will be called with a hash reference as it's first parameter, and @args as the rest of the parameters, if @args is provided.

The hash reference will have the following keys:


This is a packed sockaddr_in that specifies where the packet was received from. You can use the unpack_sockaddr_in function in Socket to unpack this.


This will be the body of the packet received.


This is the length of the packet received. It is actually the return value of recvfrom, so it will be -1 if an error occured.


This is the value of errno after recvfrom returns. This is probably only useful if len indicates there was an error.


This will do all the setup work for the UDPPump. This includes starting the backgroun thread, and registering some underlying Event::Lib events for the notification.


Returns the file handle associated with the UDPPump event.


I can't think of any reason this module wouldn't work just fine on any type of datagram socket. In fact, I imagine it'll work great on anything that recvfrom works on.

There is no way to remove a UDPPump event once it's created. This isn't implemented because I didn't care to get into the hairy mess that is required with thread cancellation, and I expect this will be used mostly in server applications where the socket is created at startup and only destroyed on exit. If you really need to be able to delete events, it'd probably not be too horrible to implement.

This has only been tested on Solaris 10. If it doesn't work for you on another platform, I'm willing to take patches or reasonably helpful error reports.


The udppump_new function is exported by default.


This module requires the excellent Event::Lib module.


Clayton O'Neill, <>

The above email address is a spamgourmet ( email address. It will stop working after it receives 20 pieces of mail. Given how badly CPAN authors get spammed, that will probably be pretty quickly. I'd recommend changing the udppump part above to anything else when sending me email.


Copyright (C) 2006 by Clayton O'Neill

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.5 or, at your option, any later version of Perl 5 you may have available.