352 CHAPTER 6 Performance Waiting time. The length of time from when a request arrives at a service until the service starts processing the request. This measure is better than turnaround time, since it captures how long the thread must wait even though it is ready to execute. The ideal waiting time is zero seconds. More sophisticated measures are also possible by combining the performance of all requests using some of these measures and some way of combining. For example, one can compute average waiting time as the average of waiting times of all requests. Similarly, one can calculate the sum of the waiting times, the variance in response time, and so on. In an interactive computer system, many requests are on behalf of a human user sitting in front of a display. Therefore, the perception of the user is another measure of the goodness of the service that a request receives. For example, an interactive user may tend to perceive a high variance in response time to be more annoying than a high mean. On the other hand, a response time that is faster than the human reaction time may not improve the perception of goodness. Sometimes a designer desires a scheduler that provides some degree of fairness, which means that each request obtains an equal share of the shared service. A sched- uler that starves a request to serve other requests is an unfair scheduler. An unfair