Skip to content
Tutorials Center
My learning

Topics

Lesson 5 of 38 · Getting Started

Exponential Service Times

Learn why queueing models use exponential service times: the memoryless property, the minimum of exponentials and the race between arrival and departure.

The M in M/M/1 stands for the exponential distribution, and it is used for service times as much as for gaps between arrivals. Why? Because it makes the math simple. The exponential distribution has a unique memoryless property: knowing how long a service has already lasted tells you nothing about how much longer it will take. That lets us describe the whole system by counting customers, without tracking elapsed times.

Definition

A service time S is exponential with rate μ when:

  • P(S > x) = e^(-μx) for x ≥ 0
  • the mean is E[S] = 1/μ
  • the variance is 1/μ², so the SCV is 1

If a teller takes 3 minutes on average, μ = 1/3 per minute. About e^(-1) ≈ 37% of services last longer than the mean, and many are much shorter, because the distribution has a heavy mass near 0.

The memoryless property

The key equation is

P(S > s + t | S > s) = P(S > t)

A service that has already run for s minutes still needs, in the probabilistic sense, a fresh exponential time. Derivation: the left side equals e^(-μ(s+t)) / e^(-μs) = e^(-μt).

Python Check the memoryless property

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Real service times are rarely exactly exponential: a task usually has some minimum duration. Exponential is a modeling choice that works as a first approximation, and we will measure how much it matters when variability changes in later chapters.

The minimum of exponentials

If two independent exponential times have rates a and b, the time until the first one finishes is exponential with rate a + b. With c busy servers each at rate μ, the time until the next departure is exponential with rate cμ. This fact drives the multi-server formulas later.

The race: arrival or departure first?

In an M/M/1 queue, from any moment when the server is busy, two clocks are running: the next arrival (rate λ) and the current service finishing (rate μ). Because both are memoryless, the probability that the arrival happens first is

P(arrival first) = λ / (λ + μ)

and the departure comes first with probability μ / (λ + μ). The time until something happens is exponential with rate λ + μ.

flowchart LR
  N["n customers"] -->|"rate λ"| U["n + 1"]
  N -->|"rate μ"| D["n - 1"]

From state n (n ≥ 1) the next event is an arrival with chance λ/(λ+μ) or a departure with chance μ/(λ+μ).

Python The race between two clocks

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Why this matters for the rest of the track

Because both clocks restart fresh at every event, the number in the system changes like a simple random walk with rates that depend only on the current count. That structure is called a birth-death process, and it is the engine behind M/M/1 and M/M/c formulas.

If your service times are far from exponential (for example nearly constant, or with rare very long jobs), M/M/1 formulas can be off by a large factor. Check the SCV first.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. A service time is exponential with mean 5 minutes. What is the rate mu per minute?

    The rate is the reciprocal of the mean: 1/5 = 0.2 per minute.
  2. 2. A service has already lasted 10 minutes (exponential). What does the memoryless property say about the remaining time?

    Memorylessness means the remaining time is distributed like a new service.
  3. 3. Arrival rate 1 per minute, service rate 4 per minute, server busy. What is the chance the next event is an arrival?

    lambda / (lambda + mu) = 1/5.
  4. 4. Three servers are all busy, each with rate 2. How is the time to the next departure distributed?

    The minimum of independent exponentials has rate equal to the sum: 2 + 2 + 2 = 6.

Summary

  • Exponential service has P(S > x) = e^(-μx), mean 1/μ and SCV 1.
  • It is memoryless, so only the count of customers needs tracking.
  • The minimum of independent exponentials is exponential with the summed rate.
  • In a busy M/M/1, the next event is an arrival with probability λ/(λ+μ).

Next: Performance Measures