Masking Latency

lastmod 2026-11-10created 2017-01-263 edits

Masking Latency refers to techniques that hide network or backend response latency and transient failures so that users do not notice them. Typical approaches include stale-while-revalidate, where a caching proxy such as Squid is placed in front and responds with an expired cache entry first while refreshing it in the background, and stale-if-error, which returns the cached copy instead when the origin server fails. At the socket level, round-trip latency is reduced by disabling the Nagle algorithm, reusing connections, and so on.

See also

• Network socket
• Network Programming
• Nagle

Favorite site

• Masking Latency & Failures with Squid ¹

References