On some variants of UNIX you may find that the performance monitoring command that has been used, e.g. vmstat, returns a disk utilization figure and the rate of transfers per second, but does not return a measured service time. If we assume that the disk is behaving according to M/M/1 queuing assumptions (often the case, so try not to worry about the detail) then we can calculate the service time from the following formula:

Service time = utilization/transfer rate.

Thus if a disk is running at 45% utilization (0.45 for the purpose of the above equation, i.e. active for 0.45 seconds each second), and has a transfer rate of 15 transfers per second, then the equation yields:

0.030 = 0.45/15
i.e. a 30 millisecond service time.

On the other hand, if we know the average queue length and transfer rate for the disk then we can again use M/M/1 assumptions to calculate the service time. For example, a disk has a measured queue length of 2. We can now use the following equation:

Queue length = utilization/(1 - utilization)
Or in other words:
utilization = 1 - (1/(Queue Length + 1))
Thus a queue length of 3 = a utilization of 0.75 or 75%.

Now that we have the utilization figure we can also calculate the service time using the previous equation.