What’s the Real Difference Between HTTP Keep‑Alive and TCP Keepalive?

During a ByteDance interview, a candidate was asked two questions: how MySQL stores NULL values and the difference between HTTP Keep‑Alive and TCP Keepalive. The article revisits these topics.

MySQL NULL storage

The first question tests understanding of MySQL record storage. In MySQL, a NULL value is represented by a bitmap in the record header; the actual column data is omitted, saving space while allowing the engine to distinguish NULL from zero or empty strings.

HTTP Keep‑Alive

HTTP Keep‑Alive is an application‑level feature that allows a single TCP connection to carry multiple HTTP request/response pairs, avoiding the overhead of establishing a new TCP connection for each request. In HTTP/1.0 it must be enabled via the Connection: Keep-Alive header; HTTP/1.1 enables it by default. The server can close the connection with Connection: close. Keep‑Alive reduces latency and enables HTTP pipelining, where a client can send several requests without waiting for each response, though the server still replies in order.

Web servers typically configure a keepalive_timeout (e.g., 60 seconds) to close idle connections and free resources.

TCP Keepalive

TCP Keepalive is a kernel‑level mechanism that periodically sends probe packets on an idle TCP connection to verify that the peer is still reachable. If the peer responds, the keepalive timer resets; if several probes fail, the connection is considered dead. Applications must enable it by setting the socket option SO_KEEPALIVE.

Summary

HTTP Keep‑Alive (application‑level) and TCP Keepalive (kernel‑level) serve different purposes: the former prolongs a TCP connection for multiple HTTP exchanges, while the latter monitors connection liveness when no data is exchanged.