tag
7 posts
Diagnose Linux TCP listener overload by separating incomplete SYN requests from established connections awaiting accept, then tune the measured bottleneck.
Operate FreeBSD's modular TCP congestion control with per-connection selection, current defaults, socket evidence, controlled benchmarks, and safe rollout criteria.
Use Haiku's BNetEndpoint with explicit TCP or UDP semantics, checked status, partial I/O loops, bounded waits, and clear socket ownership.
Set Linux TCP_USER_TIMEOUT deliberately: understand milliseconds, zero-window stalls, keepalive interaction, state scope, and application deadlines.
Operate Network.framework NWConnection streams with explicit state handling, bounded receives, framing, retries, and orderly cancellation.
A sourced history of TCP congestion control, from the 1980s collapse and Reno to SACK, NewReno, and modern CUBIC behavior.
Measure dynamic TCP port use by process and state, distinguish exhaustion from DNS or remote-service failures, and fix connection churn before widening ranges.