Latency
Overview
Latency (遲延, latency/delay) is a general concept referring to the phenomenon in which an event, stimulus, or the transmission and processing of data occurs later than the intended or expected point in time. In computer networking, it denotes the time required for a packet to travel from source to destination; in physics, it refers to a signal or wave whose phase lags behind; and in psychology, it means the time interval between a stimulus and a response. In recent years, it has been treated as a core performance metric in systems where real-time responsiveness matters, such as cloud, AI, gaming, and autonomous driving.
Main Content
Latency in Computing and Networks
Network latency is largely decomposed into four components. First, propagation delay is the physical time for a signal to travel through a medium, determined by the speed of light and distance. Second, transmission delay is the time to push a packet's bits onto a link, and is inversely proportional to packet size and bandwidth. Third, processing delay is the time a router takes to inspect headers and determine a route. Fourth, queueing delay is the time spent waiting in a buffer during congestion, and it fluctuates sharply with traffic load.
Latency must be distinguished from bandwidth. Bandwidth indicates "how much" can be sent, while latency indicates "how quickly" it arrives. In addition, jitter—the variation in latency—and packet loss have a decisive impact on real-time communication quality. For measurement, ping and RTT (Round Trip Time), along with tail latency metrics such as p95 and p99 rather than the average, are widely used.
Latency in Physics and Engineering
In signal processing, phase delay and group delay are key variables in filter design. In electrical circuits, delay lines are used to intentionally slow signals, and are employed in radar, acoustic equipment, oscilloscopes, and the like. In control engineering, system delay is a factor that undermines the stability of feedback loops, making dead-time compensation techniques important.
Latency in Psychology and Physiology
Human reaction time is formed on the order of hundreds of milliseconds, depending on the type of stimulus and level of attention. In behavioral economics, the phenomenon of "delay discounting"—valuing future rewards less than immediate ones—is used as a framework to explain impulsivity, addiction, and saving behavior.
Latency in Economics and Management
In monetary policy, a time lag exists between policy decisions and their effects on the real economy, and in supply chains, the lead time from ordering to delivery determines inventory and service levels. In software development as well, latency in the deployment pipeline reduces the speed of feedback and affects quality.
Latency in Language and Everyday Life
"지연" (jiyeon) is also commonly used as a Korean personal name, and it is widely used in everyday and legal terms such as "지연되다" (to be delayed), "지연 배상" (compensation for delay), and "지연 이자" (interest for delay).
Recent Trends
In 2024–2025, latency has once again come to the forefront. 5G's URLLC and 6G research are proceeding with the goal of sub-1 ms latency, and low Earth orbit (LEO) satellite internet has reduced satellite latency from the hundreds of milliseconds to the tens of milliseconds. With the spread of generative AI, the time to first token (TTFT) and token generation speed of LLM inference have emerged as key indicators of user experience, and edge computing and on-device AI are drawing attention as strategies to reduce cloud round-trip latency. In cloud gaming, XR, remote robot control, and autonomous driving V2X, p99 latency and jitter management are being included in regulatory discussions as safety requirements.
Related Topics
- [[Latency]]
- [[Bandwidth]]
- [[Edge Computing]]
- [[5G]]
- [[Jitter]]