ITCE320 - Ch6: Server Architecture and Multithreading

Single-Thread Programs

• A single-thread program is a script with statements and expressions executed in sequence
• Easier to design and used in programs that handle single connection/session at a time
• Vulnerable to denial-of-service attacks
• Connect and keep the connection on, keeping the server blocked at recv() expression
• Resend some data before the blockage timeout (if it is configured)

Behavior with Multiple Connections

• When a client tries to connect to a single-threaded server that is in session with another client
• The OS starts a three-way TCP handshake with the second client to set up a connection
• The connection will be kept at OS’s listen queue until the ongoing conversation is concluded
• Once the server code has looped back to its next call to accept(), the second client’s connection will be available to the server
• Start exchanging requests and responses over that socket

Single-Threaded Server and CPU Utilization

• Poor resources utilization
• recv() causes blockage, and no other expressions are executed during blockage time
• Until receive data from the client
• Timeout

Single-Thread Server Example

• Create a socket and bind it to the loopback address and port 22221
• Listen for incoming connections
• Accept a connection and receive client name
• Start a chat loop to send and receive messages
• Close the connection when the client sends "quit"

Multi-process and Multi-Thread Programs

• Multiple lightweight flow of executions running at the same time
• Multi-threading: technique that allows an execution of several tasks of one process at the same time
• Single core: execution of different codes are all queued, no parallel processing for both single and multi-threaded scripts
• Multiple cores: parallel executions at the same time

Multi-process vs Multi-threaded

• Multi-process Programming: each process runs as a standalone program, allocated code and data memory for every process
• Multi-threaded: multiple threads running in a single process, sharing code and data memory
• Use join() method to block until a thread terminates

Returning Threads

• The Main Thread: return the main Thread object
• The Current Thread: return the current Thread object
• Returning list of threads: use threading.enumerate() to get a list of active threads

Example: Wait for all Active Non-Main Threads

• Get a list of active threads using threading.enumerate()
• Iterate through the list and join each thread

Multithreaded Server

• Create multiple threads to handle multiple client connections
• Each thread handles a separate client connection
• Use threading and socket libraries to create a multithreaded server

Background

• The Internet is used to share services
• Entities to provide services need to be available, approachable, reliable, and scalable
• Connected by program entities to ask for services

Server

• A server provides a service that is needed by one or more clients
• Can be a single server or a server farm (mirrored/cloud)
• Example: Apache web server responds to requests from Firefox as the web client

Reaching Services

• Requests are generated by clients
• Clients can be at the same host as the server or at a remote machine
• Reaching the server mechanisms:
  • Common servers use well-known common port numbers
  • Not common servers need a mechanism to verify the server port number to the client