Skip to content
Tutorials Center
My learning

Topics

Lesson 18 of 38 · Multi-Server Queues

One Line or Many? The Power of Pooling

Compare c separate M/M/1 lines with one shared M/M/c line and one fast server. See why pooling cuts waiting and what the limits are.

You walk into a bank with four tellers. Some banks give each teller their own line; others have one snake-shaped line feeding the next free teller. Does it matter? A lot. This lesson compares three designs with the same total capacity and the same total customer stream.

Three designs

Let c servers each work at rate μ, and let the total arrival rate be λ.

  1. Separate lines. Customers pick a teller at random (or arrive evenly), so each teller sees an M/M/1 queue with arrival rate λ/c. Time in system: W = 1 / (μ - λ/c).
  2. One shared line (pooled). An M/M/c queue with all c servers. Time in system: W = C/(cμ - λ) + 1/μ.
  3. One fast server. Merge the servers into a single one that works c times as fast, rate cμ. An M/M/1 queue with W = 1 / (cμ - λ).
flowchart LR
  subgraph A["Separate lines"]
    A1["line"] --> S1["server"]
    A2["line"] --> S2["server"]
  end
  subgraph B["Pooled"]
    B1["one line"] --> T1["server"]
    B1 --> T2["server"]
  end

Same servers, different line structure. In the pooled design nobody waits while another server is idle.

The utilization is the same in all designs: ρ = λ/(cμ). Only the arrangement differs.

Why separate lines lose

With separate lines, one teller can be idle while another has a long line. Capacity is wasted on average, and customers stuck behind a long service wait longer. Pooling removes this imbalance: as soon as any server is free, the next customer goes to it.

The numbers

We fix ρ = 0.8 and μ = 1 per minute, and let c grow. The code prints the average time in system for each design.

Python Three designs compared

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.

For c = 1 all three coincide. As c grows, separate lines always cost 5 minutes (the M/M/1 value at ρ = 0.8), while the pooled system improves steadily. The "one fast server" design is faster still, because it shortens service itself: a customer's own service time drops from 1 minute to 1/c.

Waiting only

If we measure only the queueing delay Wq, the ranking changes, because the fast server's advantage was partly shorter service, not shorter waiting. Print the mean waiting before service starts:

Python Waiting before service starts

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.

In this comparison the pooled design has the shortest waiting before service for c > 1, while the fast server has the shortest total time. Which is best depends on what the customer cares about: waiting in line (frustration) or total time.

Limits of pooling

Pooling is not free. Real examples where it fails or needs care:

  • Specialization. If servers have different skills, a shared line may send a customer to someone who cannot help. Call centers use skill-based routing for this reason.
  • Travel or switching costs. Pooling several locations means customers travel farther or workers change tasks.
  • Priority needs. If some customers are more urgent, a single FIFO line ignores that.
  • Variance matters. The analysis assumes exponential service. With very variable jobs, pooling still helps, but the numbers differ.

Rule of thumb: when jobs are interchangeable and servers are identical, use one shared line. The gain grows with the number of servers.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. With c separate M/M/1 lines, each line sees what arrival rate?

    The total arrival stream is split evenly among the c lines.
  2. 2. Why does a shared line beat separate lines at the same utilization?

    A pooled queue never leaves a server idle when someone is waiting, so capacity is used better.
  3. 3. For c = 1, which statement is true?

    With one server there is nothing to pool or merge.
  4. 4. Which situation is a reason NOT to pool?

    If servers differ in skills, routing everyone into one line may mismatch customers and servers.

Summary

  • Compare c separate M/M/1 lines, one M/M/c line and one c-times-faster server at equal load.
  • Pooling beats separate lines by avoiding idle servers next to long lines.
  • A single fast server reduces service time too, so it wins on total time.
  • Pooling has limits: specialization, travel, priorities.

Next: The M/M/∞ Queue