Sections 7.1 and 7.2 have discussed two sets of theoretical problems that arise in simulations of urban services. In this and the next section we shall turn to two questions that always arise in connection with the actual preparation of simulation programs. We shall first address the problem of "advancing the clock" in a simulation.

In terms of simulating the passage of time, there are two basic types of simulations: (1) fixed-time-increment simulations, and (2) event-paced simulations. In fixed-time-increment simulations, time is advanced by regular intervals, say, every 8 seconds or every hour or every 3 years. The length of the fixed interval is, naturally, predetermined by the programmer and is chosen on the basis of the context of the simulation. In a macroscopic simulation of the future of the economy of a large city, for instance, the fixed time increment might reasonably be chosen as I year. The simulation would then follow from year to year the changes in the various economic activities in that city. On the other hand, a simulation concerned with producing estimates of the average response time by a local police department to emergency calls might use a fixed time interval of, say, I minute or of 30 seconds or lessdepending on the amount of accuracy desired. In general, the length of these fixed time intervals are of the same order of magnitude as what might be called loosely the "natural time units" of the simulated system.

The concept of an event-paced simulation is less obvious than that of fixed time increments. It is an extremely useful one, however, and is used at least as often as the fixed-time-increment approach-especially in the case of the simulation of urban service systems. Basically in this case, the clock is advanced only to those instants in time when certain important (for the simulated system) events take place. Such events might be the arrival or departure of customers in a queueing system, or the occurrence of a fire alarm in the simulation of a fire department. Obviously, in event-paced simulations, time is advanced in irregular intervals.

No general guidelines can be offered regarding the choice between fixedtime-increment and event-paced simulation. This choice depends on the requirements of the problem. Clearly, however, an event-paced simulation provides a maximum amount of detail about activities during the simulated period and does not waste any effort dealing with times at which "nothing occurs."