Discrete Event Simulation

One of the most important applications of priority queues is in discrete event simulation. Simulation is a tool which is used to study the behavior of complex systems. The first step in simulation is modeling. We construct a mathematical model of the system we wish to study. Then we write a computer program to evaluate the model. In a sense the behavior of the computer program mimics the system we are studying.

The systems studied using discrete event simulation  have the following characteristics: The system has a state  which evolves or changes with time. Changes in state occur at distinct points in simulation time. A state change moves the system from one state to another instantaneously. State changes are called events.

For example, suppose we wish to study the service received by customers in a bank. Suppose a single teller is serving customers. If the teller is not busy when a customer arrives at the bank, the that customer is immediately served. On the other hand, if the teller is busy when another customer arrives, that customer joins a queue and waits to be served.

We can model this system as a discrete event process as shown in Figure . The state of the system is characterized by the state of the server (the teller), which is either busy or idle, and by the number of customers in the queue. The events which cause state changes are the arrival of a customer and the departure of a customer.

  
Figure: A simple queueing system.

If the server is idle when a customer arrives, the server immediately begins to serve the customer and therefore changes its state to busy. If the server is busy when a customer arrives, that customer joins the queue.

When the server finishes serving the customer, that customer departs. If the queue is not empty, the server immediately commences serving the next customer. Otherwise, the server becomes idle.

How do we keep track of which event to simulate next? Each event (arrival or departure) occurs at a discrete point in simulation time  . In order to ensure that the simulation program is correct, it must compute the events in order. This is called the causality constraint--events cannot change the past.

In our model, when the server begins to serve a customer we can compute the departure time of that customer. So, when a customer arrives at the server we schedule an event in the future which corresponds to the departure of that customer. In order to ensure that events are processed in order, we keep them in a priority queue in which the time of the event is its priority. Since we always process the pending event with the smallest time next and since an event can schedule new events only in the future, the causality constraint will not be violated.

Copyright © 1998 by Bruno R. Preiss, P.Eng. All rights reserved.