Skip to content
Tutorials Center
My learning

Topics

Lesson 19 of 38 Ā· Multi-Server Queues

The M/M/āˆž Queue: Infinite Servers

Study the M/M/infinity queue where nobody waits. The number in the system is Poisson, which gives a simple way to size capacity for self-service systems.

Some systems behave as if every customer gets a server immediately: a self-service website, a parking lot, people browsing a store, simultaneous phone conversations on a big network. We model them with infinitely many servers, the M/M/āˆž queue. Nobody ever waits; the interesting question is how many are in the system at once.

The distribution

M/M/āˆž is a birth-death process with Ī»n = Ī» and μn = nμ, because each of the n customers present departs at rate μ. The product form gives

Pn = P0 Ɨ Ī»^n / (n! μ^n) = P0 a^n / n!

with a = λ/μ. Normalizing, P0 = e^(-a), so

Pn = e^(-a) a^n / n!

The number of customers in an M/M/āˆž system is Poisson with mean a. There is no stability condition: for any Ī» and μ the system settles down, since there are always enough servers.

Measures:

Measure Value
L (mean number present) a = λ/μ
Lq 0
Wq 0
W 1/μ
Variance of the number present a

Little's law checks out: L = Ī» Ɨ (1/μ) = a.

Sanity check by simulation

Customers arrive as a Poisson process, each stays for an exponential time, and nobody waits. At a given time t, the number present is the number of arrivals up to t minus the number of departures up to t. We sample at many times, after the system has warmed up, and compare to Poisson(a).

Python Occupancy is Poisson

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

Program input Ā· 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.

The mean and variance both come out near 12, as Poisson requires.

Capacity planning

M/M/āˆž is not just a toy. If a system has m seats, connections or lines, but a request that finds all of them busy is not simply delayed (it might fail or you might add capacity), then M/M/āˆž tells you how many resources are occasionally needed. The question "How large must m be so that occupancy exceeds m less than 1% of the time?" is a Poisson tail calculation: find the smallest m with P(N > m) ≤ 0.01.

Because N is Poisson with mean a, the answer is about a plus a few multiples of √a, as the normal approximation for a Poisson variable suggests.

Python How many seats?

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

Program input Ā· 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.

As a grows, the safety margin above the average grows roughly like √a, so large systems need relatively less spare capacity. We will see the same square-root behavior for waiting systems in the next lesson.

M/M/āˆž results (Poisson occupancy, W = 1/μ) hold even when service times are not exponential, as long as arrivals are Poisson. This makes it one of the few robust results in queueing theory.

Check your understanding

Quick check Ā· 4 questions

of 4 answered You scored / 4 Ā· perfect!

  1. 1. In an M/M/infinity queue with lambda = 10 per hour and mu = 2 per hour, what is the mean number present?

    L = a = lambda / mu = 5.
  2. 2. What is the distribution of the number in an M/M/infinity system?

    The product form gives Pn = e^(-a) a^n / n!.
  3. 3. What is the mean waiting time in the queue Wq?

    Every customer finds a free server immediately, so there is no queueing.
  4. 4. What is the stability condition for M/M/infinity?

    With unlimited servers the system always reaches a steady state.

Summary

  • M/M/āˆž: no waiting; each customer has their own server.
  • The number present is Poisson with mean a = Ī»/μ; W = 1/μ, Lq = Wq = 0.
  • It is always stable and the result is insensitive to the service distribution.
  • Poisson tails size resources: spare capacity grows like √a.

Next: Staffing a Call Center