Short vs Long TCP Connections: When and Why to Use Each

1. TCP Connection

When using TCP, a connection must be established before any read/write operation and released after communication ends. Establishment requires a three‑way handshake, while termination requires a four‑step handshake, both consuming resources and time.

Classic three‑way handshake diagram:

Classic four‑step termination diagram:

TCP state diagram (ten states):

Note: When one side receives a FIN, the kernel makes read return 0 to notify the application that the peer has terminated data transmission; sending FIN usually results from the application closing the socket.

TCP 2MSL problem diagram:

2. TCP Short Connection

In a short‑connection scenario, the client initiates a connection, sends a message, receives a response, and then either side can close the connection, typically the client. Short connections are simple to manage and only exist while useful.

3. TCP Long Connection

In a long‑connection scenario, after the initial handshake, the client and server keep the connection open for multiple read/write operations. TCP keep‑alive can be used: if no activity occurs for a configured period (e.g., two hours), the server sends probe packets; after several unanswered probes, the connection is considered dead and closed.

Long connections can lead to resource exhaustion if many idle connections accumulate, so servers may close long‑idle connections or limit the maximum number of concurrent long connections per client.

What are “long connection” and “short connection”?

Long connection: a TCP socket remains open regardless of data transfer, offering lower latency for frequent communication but consuming more resources. Short connection: the socket is opened, data is exchanged, and then immediately closed, freeing resources quickly.

HTTP Protocol Long and Short Connections

Short connection: client establishes a connection, sends a request, receives a response, and closes the connection. Long connection (persistent HTTP): the client and server keep the TCP connection open for multiple requests/responses, reducing handshake overhead.

When to use long vs short connections

Use long connections for frequent, point‑to‑point communication where the number of simultaneous connections is limited (e.g., database connections). Use short connections for high‑concurrency scenarios like web servers, where each client performs brief requests and keeping many connections open would strain resources.

Sending and receiving methods

Asynchronous: sending and receiving are independent, can be full‑duplex or half‑duplex.

Synchronous: send a message and wait for a response, typically with a timeout.

Message formats and I/O models

Non‑blocking I/O: read attempts return immediately with a timeout if no data is available.

Blocking I/O: read blocks until data arrives.

Looped read/write: handle partial reads/writes by looping until the full message is transferred, optionally using a length header for framing.