Interarrival time is the duration between two consecutive arrivals in a system. If three customers walk into a coffee shop at 9:00, 9:04, and 9:11, the interarrival times are 4 minutes and 7 minutes. It’s a foundational concept in queuing theory, networking, and operations research, used to predict how long lines get, how servers handle load, and how systems perform under stress.
How Interarrival Time Works
The idea is straightforward: you pick a point where things arrive (a checkout counter, a web server, an emergency room) and measure the gap between each arrival. Those gaps are your interarrival times. They can be fixed, like a factory machine that produces one widget every 30 seconds, or random, like patients showing up at a clinic throughout the day.
Interarrival time has a simple inverse relationship with arrival rate. If the average interarrival time is 5 minutes, you can expect about 12 arrivals per hour. Expressed as a formula: the arrival rate (λ) equals 1 divided by the mean interarrival time. So if customers arrive on average every 2 minutes, λ = 0.5 customers per minute, or 30 per hour. This relationship lets you move freely between “how often do things show up” and “how long between each one.”
Why It Usually Follows an Exponential Distribution
In most real-world systems where arrivals happen independently of one another (customers don’t coordinate when they show up, network packets don’t synchronize), interarrival times follow an exponential distribution. This is tied directly to the Poisson process, which is the standard model for random arrivals. If events occur at a constant average rate and each arrival is independent of the last, the time between arrivals is exponentially distributed with a mean of 1/λ.
The exponential distribution has a key property called “memorylessness.” If you’ve already waited 3 minutes without an arrival, the expected remaining wait is the same as it was when you started. The system has no memory of how long it’s been. This feels counterintuitive, but it’s what makes random arrivals genuinely random: the chance of someone arriving in the next 10 seconds doesn’t change based on how long you’ve been waiting.
Not all systems behave this way. A bus that runs on a schedule has deterministic (fixed) interarrival times. A system that batches arrivals together, or one where arrivals tend to cluster, might follow a different distribution entirely. But the exponential model is the default starting point because it accurately describes a wide range of scenarios, from customer arrivals to radioactive decay.
Interarrival Time and Queue Performance
What makes interarrival time so useful is that it directly controls how congested a system gets. Two systems can have the same average arrival rate but wildly different queue lengths depending on how variable the interarrival times are. If customers arrive almost exactly every 5 minutes, a server that takes 4 minutes per customer handles the load easily. But if customers sometimes arrive in bursts (two in 30 seconds, then nobody for 15 minutes), the same server will develop long lines during those bursts even though the average rate hasn’t changed.
This variability is measured using the coefficient of variation, which compares how spread out the interarrival times are relative to their average. For exponential interarrival times, this value is 1. For more regular, predictable arrivals, it drops below 1. Higher variability means more congestion, longer waits, and less predictable service. This is why scheduling systems (appointments, reservations, timed entry) work so well: they reduce interarrival time variability, which smooths out demand and shrinks wait times even without adding capacity.
Applications in Networking
In computer networks, interarrival time describes the gap between consecutive packets reaching a destination. When a voice or video call sends packets, they’re typically generated at equal intervals. But as those packets travel through routers and compete with other traffic, the spacing between them becomes uneven. This variation in inter-packet arrival time is called jitter, and it’s one of the main causes of choppy audio or frozen video in real-time calls.
Jitter gets worse when the network is congested, because packets spend unpredictable amounts of time waiting in router queues. The distribution of inter-packet arrival times tends to skew toward longer gaps during heavy traffic, with a hard lower bound (packets can’t arrive faster than the network physically allows). Research suggests these arrival times roughly follow a bell-curve distribution, though skewed. Applications handle jitter by using a buffer that collects packets and plays them back at a steady rate, trading a small amount of added delay for smoother playback.
Applications in Computing and Simulation
Operating systems and simulation tools use interarrival times to model workloads. In CPU scheduling simulations, for instance, jobs arrive with specified interarrival times to test how different scheduling algorithms handle varying loads. A typical simulation might define 1,000 jobs with a mean interarrival time of 120 milliseconds and a mean processing time of 100 milliseconds. Because the jobs arrive slightly faster than they can be processed on average, a queue builds up, and the simulation reveals how different scheduling strategies manage that backlog.
Cloud computing platforms use the same concept to plan capacity. By measuring the interarrival times of user requests, engineers can model peak loads, estimate how many servers are needed, and set autoscaling thresholds. If request interarrival times suddenly shrink (meaning traffic is spiking), the system spins up additional resources. The entire discipline of capacity planning rests on understanding arrival patterns, and interarrival time is the most basic unit of that understanding.
How to Calculate It
If you have raw arrival data (timestamps of when each event occurred), calculating interarrival times is simple: subtract each arrival time from the next one. For a dataset of 100 arrivals, you get 99 interarrival times. From there, you can compute the mean, look at the distribution, and estimate the arrival rate.
If you only know the arrival rate, you can work backward. An arrival rate of 10 per hour gives a mean interarrival time of 6 minutes (1 divided by 10, converted to minutes). If you’re assuming a Poisson process, the individual interarrival times are exponentially distributed around that mean, which means some gaps will be very short and others quite long. About 63% of interarrival times will be shorter than the mean, and a few will be several times longer. That lopsided spread is a natural feature of exponential distributions and matches what you’d observe watching random arrivals in real life.