Encyclopedia of Computer Science and Technology

400 recursionand with the potential consequences of failure. An air trafficcontrol system may be able to take a few seconds betweenprocessing radar samples, but it better get it right in time.Systems like this where real-time response is absolutelycrucial are sometimes called “hard real-time systems.”Other systems are less critical. A streaming audio systemhas to keep its buffer full so it can play in real time,but if it stutters once in a while, no one’s life is in danger.(And since download rates over the Internet can vary formany reasons it’s not realistic to expect too perfect a levelof performance.) A slower “soft real-time system” like abank’s ATM system should be able to respond in tens of seconds,but if it doesn’t, the consequences are mainly potentialloss of customers and revenue. A fairly wide variation inresponse time may be acceptable as long waits don’t occuroften enough to drive away too many customers.To put together the system, the engineer must look atthe inherent speed of the sampling device (such as radar,camera, or simply the keyboard buffer). The speed of theprocessor(s) and the time it takes to move data to and frommemory are also important. The structure, strengths, andweaknesses of the host operating system can also be a factor.Some operating systems (including some versions ofUNIX) feature a guaranteed maximum response time forvarious operating system services. This can be used to helpcalculate the “worst case scenario”—that combination ofinputs and the existing state of the system that should resultin the slowest response.Another approach available in most operating systems isto assign priority to parts of the processing so that the mostcritical situations are guaranteed to receive the attention ofthe system. However, things must be carefully tuned so thateven lower priority tasks are accomplished in an acceptablelength of time.The design of the data structure or database used tohold information about the process being monitored is alsoimportant. In most databases the age of the data is not thatimportant. For a payroll system, for example, it might besufficient to run a program once in a while to weed outpeople who are no longer employees. For a nuclear powerplant, if data is getting too old such that it’s not keepingup with current condition, some sort of alarm or evenautomatic shutdown might be in order. With a system thathas softer constraints (such as an automatic stock tradingsystem), it may be enough to be able to get most trades donewithin a specified time and to gather data about the performanceof the system so the operators can decide whether itneeds improvement.Real-time systems are increasingly important becauseof the importance of the activities (such as air traffic controland power grids) entrusted to them, and because oftheir pervasive application in everything from cars to cellphones to medical monitors (see embedded system andmedical applications of computers). The systems alsotend to be increasingly complex because of the increasinginterconnection of systems. For example, many real-timesystems have to interact with the Internet, with communicationsservices, and with ever more sophisticated multimediadisplay systems. Further, many real-time systems mustuse multiple processors (see multiprocessing), which canincrease the robustness and reliability of the system butalso the complexity of its architecture, and thus the difficultyin determining and ensuring reliability.Further ReadingButtazzo, Giorgio C. Hard Real-Time Computing Systems: PredictableScheduling Algorithms and Applications. 2nd ed. NewYork: Springer, 2005.Cheng, Albert M. K. Real-Time Systems: Scheduling, Analysis, andVerification. Hoboken, N.J.: Wiley, 2002.Resources for Real-Time Computing. TechRepublic. Availableonline. URL: http://search.techrepublic.com.com/search/real-time+computing.html. Accessed August 19, 2007.recursionEven beginning programmers are familiar with the ideathat a series of program statements can be executed repeatedlyas long as (or until) some condition is met (see loop).For example, consider this simple function in Pascal. Itcalculates the factorial of an integer, which is equal to theproduct of all the integers from 1 to the number. Thus factorial5, or 5! = 1 * 2 * 3 * 4 * 5 = 120.Function Factorial (n: integer) : integerBegini: integer;For i = 1 to n doFactorial := Factorial * i;End.If the main program has the line:Writelin (Factorial (5));then the 5 is sent to the function, where the loop simplymultiplies the numbers from 1 to 5 and returns 120.However, it is also possible to have a function call itselfrepeatedly until a specified condition is met. This is calledIn recursion, a procedure calls itself until some defined conditionis met. In this example of a Factorial procedure, F(1) is defined toreturn 1. Once it does, the returned value is plugged into its caller,which then returns the value of 1 * 2 to its caller, and so on.