22.6 Thread-per-request and Process-per-request Implementations

22.6 Thread-per-request and Process-per-request Implementations

This section specifies programs to compare the performance of thread-per-request and process-per-request server implementations when disk I/O is not a factor. Write two server programs, thread_per_request and process_per_request , that are to be tested under the same offered load.

The thread_per_request server takes the port number on which to accept connections as a command-line argument. The main thread listens for connection requests and creates a detached thread to handle the communication. The detached thread is passed an array containing the communication file descriptor and a thread number. The thread calls handle_request of Section 22.4.2 and then exits.

Implement a program that uses child processes instead of threads to handle the requests. The process_per_request program is similar to thread_per_request except that the main program waits for completed children (e.g., Example 3.13). Be sure to use the WNOHANG option when waiting so that the server can process concurrent children. Compare the performance of these two approaches as a function of the offered load. Present your results with graphics and a written discussion.