This is a simple study to try several types of server architetures.


There are currently 3 types of servers:

- One thread per connection (one-thread-connection.c)
  There's a main loop blocking on accept() and one thread is created to
  handle each client

- Single thread, using libevent (libevent-single-thread.c)
  Make use of libevent, to do asynchronous IO

- libevent + pool of threads (pool-threads-libevent.c)
  Uses asynchronous IO but leaves the actual IO job to a separated thread,
  previously created (worker-thread)

There's also a single client (client.c) that can be used to connect to
all those servers.

The idea here is to make experiments on these types of servers and help to
define which architecture to use on your server implementation. All of them
seem to have pros and cons. The winner probably will depend on your requirements.


To compile:

  ./autogen.sh
  ./configure
  make
  (no need to make install)

In the src/ directory you'll find the binaries.

Have fun!