Latency (engineering)

Latency is the time between an input or request and the corresponding response or completed operation. It includes delays introduced by processing, communication, storage, and queueing.

Round-trip time: The elapsed time for a signal or request to travel to a destination and for a response to return. It measures network delay across a complete request-and-response exchange.
Throughput: The amount of work a system completes per unit time. A system can process many requests per second while each request still waits a long time.
Latency compensation: Techniques that reduce or conceal the effects of delay in an interactive or control system. Prediction and buffering can preserve responsiveness when delay cannot be removed.
Latency versus throughput: The distinction between time per operation and the rate of completed operations. Optimizing batch size or concurrency can improve throughput while increasing individual response time.
Queueing delay: The time an item waits in a queue before receiving service. Congestion adds waiting time before processing or transmission can begin.
Bandwidth: The maximum rate at which a communication channel can carry data. Bandwidth limits transfer capacity, while latency measures elapsed time.
Real-time computing: Computing in which correctness depends on producing results within specified time constraints. Latency matters when a late result can be as harmful as an incorrect one.
Caching: Storing copies of data so later requests can retrieve them without repeating the original computation or retrieval. A cache can shorten response time, but stale data and misses complicate the tradeoff.
Propagation delay: The time a signal takes to travel through a physical medium between two points. Distance and signal speed impose a transmission delay that processing cannot eliminate.
Latency measurement: The practice of recording elapsed time between defined events in a system. Consistent start and end points are essential for meaningful latency figures.

Linked from 8 pages