Keep-Alive Header

What is Keep Alive?

• In the HTTP protocol, it allows multiple requests and responses between the client and server to be processed through a single TCP connection

  • HTTP operates by establishing a new TCP connection between the client and server for each request and closing the connection after receiving the response

    • This causes overhead from establishing and terminating connections for each request, which can lead to performance degradation when many requests need to be processed

  • Keep-Alive is a mechanism introduced to solve this problem, maintaining the connection between the client and server and allowing multiple requests and responses to be processed over the same connection

  • When the client includes a "Connection: keep-alive" header in a request, the server can continue sending responses for multiple requests without closing the connection

• Starting from HTTP/1.1, the Keep-Alive feature, which reuses already established TCP connections, is supported by default

• Since the 3-way handshake process is skipped, performance improvement can be expected

  • Reduces overhead from connection setup and teardown and improves performance by reducing network latency

Keep Alive Characteristics

• The Keep-Alive retention time is a structure that maintains the session for a defined duration from the last I/O access termination on the connected socket, even if there is no access

  • In other words, if access occurs within the defined time, the connected state can be maintained

• To use the Keep-Alive header, the Connection header must be set to keep-alive

  • However, in HTTP/2 it is ignored, and connection management is handled by different mechanisms within that protocol

• Event-driven architectures that use non-blocking, such as Nginx, do not occupy Threads while using Keep-Alive, making them advantageous for concurrent processing

Why Set a Keep-Alive Timeout

• Server resources are not infinite

• Maintaining connections indefinitely causes losses to the server

• In other words, the number of sockets that can establish connections with the server is limited, and if connections are maintained for too long, it can lead to situations where others cannot connect

Keep-Alive in HTTP/1.0+ and HTTP/1.1

• In 1.0+, configuration was needed to use Keep-Alive,

• In 1.1, configuration is needed to disconnect Keep-Alive

• In other words, the purpose of the configuration differs: whether it's for using it or for disconnecting it!