STA 247, Assignment 3 - output and discussion. Here is the output of five runs of the program, with different random number seeds. (See also the plots, from the last of these runs.) > do.stuff(1) Rates of requests being satisfied: elevator = 0.107 , closest track = 0.161 Average wait times: elevator = 46.615 , closest track = 30.872 > do.stuff(2) Rates of requests being satisfied: elevator = 0.107 , closest track = 0.158 Average wait times: elevator = 46.668 , closest track = 31.598 > do.stuff(3) Rates of requests being satisfied: elevator = 0.108 , closest track = 0.164 Average wait times: elevator = 46.273 , closest track = 30.386 > do.stuff(4) Rates of requests being satisfied: elevator = 0.107 , closest track = 0.161 Average wait times: elevator = 46.817 , closest track = 30.748 > do.stuff(5) Rates of requests being satisfied: elevator = 0.105 , closest track = 0.16 Average wait times: elevator = 47.318 , closest track = 31.024 One can see that the elevator scheme has a lower rate of satisfying requests, and a longer average wait time. The accuracy of the estimates is pretty good, since they vary little from one run to another, and is certainly good enough to be sure of these conclusions. However, the closest track scheme sometimes leads to a process having to wait a very long time for its request to be satisfied, as can be seen from the histograms (up to 643ms for the closest track scheme, as opposed to 137ms for the elevator scheme). Something like the closest track scheme is probably best for most purposes, especially if it is modified a bit bit to avoid very long wait times. If limiting the maximum wait time is a big issue, however, the elevator scheme might be best, since one can figure out that it will never take more than 5 + 19*(2+5) = 138ms to satisfy a request.