keepalive

Keepalive is a technique in computer networking used to maintain active connections and detect unresponsive peers by periodically sending control messages or probes. It helps ensure that a connection through unreliable networks, NAT devices, or idle proxies remains usable, and it helps identify when a peer has become unavailable.

TCP keepalive operates at the transport layer. When a connection is idle, the operating system may send a keepalive packet after a configured idle period. If the peer responds, the timer resets; if responses fail after a number of probes, the connection is considered dead. Settings such as idle time, probe interval, and maximum probes vary by operating system. Enabling TCP keepalive can aid failure detection but adds network traffic and may be ignored by middleboxes or firewalls.

HTTP keep-alive, or persistent connections, allows a single TCP connection to be reused for multiple HTTP requests. In HTTP/1.1, connections are typically persistent by default, reducing latency from establishing new TCP connections. Server and client may still impose idle timeouts and connection limits, which can terminate otherwise idle connections.

Application-layer keepalives are common in many protocols. WebSocket connections use ping/pong frames to verify liveness. MQTT defines a keepalive parameter that requires communication at least every specified interval. SSH can employ keepalive messages to prevent disconnects in idle sessions. These mechanisms are often more reliable within firewalls and NATs than TCP keepalives alone.

Considerations include balancing timely failure detection with overhead and compatibility with network devices. Keepalive settings interact with proxies and NATs and should align with the protocol’s semantics; in many deployments, application-level heartbeats or protocol pings are preferred over aggressive TCP keepalives.