Waiting Line Analysis and Queuing Systems

Common System Configurations

Queuing systems are often categorized by the number of servers and the number of service phases they involve:

• Single Server, Single Phase: One cashier at a small convenience store.
• Single Server, Two Phases: A patient at a clinic who must first register and then see a nurse.
• Multiple Server, Single Phase: Several cashiers in a grocery store where customers choose a line.
• Multiple Server, Multiple Phases: An airport passenger processing system involving check-in, security, and boarding.

Mathematical Foundations: Arrival and Service Rates

To model these systems mathematically, we rely on probability distributions to describe how customers arrive and how long they take to be served.

The Poisson Distribution (Arrivals)

The arrival rate ($\lambda$) is the number of units arriving per period. When arrivals are random and independent, we use the Poisson distribution to determine the probability of a specific number of arrivals ($n$) occurring within a time period ($T$):

$P_n = \frac{(\lambda T)^n e^{-\lambda T}}{n!}$

The Exponential Distribution (Service)

The service rate ($\mu$) represents the capacity of the server (customers per unit of time). The time a customer spends being served is often modeled using the exponential distribution:

$P_{t \le T} = 1 - e^{-\mu T}$

Exam Tip: Always ensure your units for $\lambda$ (arrivals per hour) and $\mu$ (service per hour) match the time $T$ you are analyzing. If a problem gives you service time in minutes but an arrival rate in hours, convert the minutes to hours before calculating.

Capacity Planning and the Economic Trade-off

One of the most critical roles of an operations manager is capacity planning: deciding how much service capacity to invest in. This involves a fundamental capacity-waiting trade-off:

1. Service Capacity Costs: Increasing capacity (e.g., hiring more workers or adding more counters) increases the cost of running the system.
2. Customer Waiting Costs: Longer waits lead to lost business, loss of goodwill, or even physical costs like paying employees to wait for a machine.

The goal of queuing analysis is to find the optimum cost, which is the point where the sum of these two costs is minimized.

Performance Measures and Little's Law

To evaluate a system, we use several key performance measures:

• Server Utilization ($\rho$): The proportion of time a server is busy.
• $L_s$ and $L_q$: The average number of customers in the system (including those being served) and in the waiting line, respectively.
• $W_s$ and $W_q$: The average time a customer spends in the system and the average time spent waiting in line.

A fundamental relationship used to link these measures is Little's Law:

$L = \lambda W$

In the context of a queue, this can be expressed as: $L_s = \lambda W_s$ $L_q = \lambda W_q$

Modeling the Single-Server System

In a standard single-server, single-phase model with an infinite customer population and FCFS discipline, we can derive several specific formulas:

• Utilization: $\rho = \frac{\lambda}{\mu}$
• Probability of zero customers in the system: $P_0 = 1 - \rho$
• Average number of customers in the system: $L_s = \frac{\lambda^2}{\mu(\mu - \lambda)} + \frac{\lambda}{\mu}$ (or more simply $L_s = L_q + \rho$)

Worked Example: The Bakery

Suppose a bakery has customers arriving at $\lambda = 18$ per hour (Poisson) and a server who can handle $\mu = 15$ customers per hour (exponential).

Wait! In this specific scenario, $\lambda > \mu$ (18 > 15). In a stable queuing system, the service rate must exceed the arrival rate ($\mu > \lambda$), otherwise the line will grow infinitely. Let's look at a valid example from the text:

A clerk serves customers at $\mu = 35$ per hour, and $\lambda = 30$ per hour:

1. Utilization: $\rho = \frac{30}{35} = 0.857$ (The clerk is busy 85.7% of the time).
2. Average number in line ($L_q$): Using the formula $L_q = \frac{\lambda^2}{\mu(\mu - \lambda)}$, we get $L_q = 5.14$ customers.
3. Average time in line ($W_q$): $W_q = \frac{L_q}{\lambda} = \frac{5.14}{30} = 0.171$ hours, or approximately 10.2 minutes.

Limitations of the Models

While these mathematical models are powerful, they have limitations. They often assume:

• Arrivals follow a perfect Poisson distribution.
• Service times follow a perfect exponential distribution.
• The customer population is infinite and "patient" (they won't leave the line).

In reality, customers might "balk" (refuse to join a long line) or "renege" (leave the line after waiting too long). For more complex systems involving non-standard distributions or complex human behaviors, managers often turn to simulation rather than pure mathematical formulas.