Latency is a shape, not a number
A single average hides the interesting part. Looking at variation over time often explains behaviour that a headline metric cannot.
Short technical notes about networks, protocols, failure modes and the useful details discovered while building small systems.
A single average hides the interesting part. Looking at variation over time often explains behaviour that a headline metric cannot.
Systems become easier to reason about when one broken component has a clearly limited effect on everything around it.
When a connection behaves unexpectedly, tracing each boundary in reverse can be more useful than staring at the endpoint.
Quiet defaults accumulate into system behaviour. Writing them down is often the first step toward understanding the whole.
The useful question is rarely “does it work?” but “how does it fail?”