When a job is running, we know that at some time in the future it will complete and free all of its processors. Given the age of the job, we can use the CLM to calculate the probability that it will have completed before time t.
We approximate this behavior by a model in which processors are a continuous (rather than discrete) resource, which jobs release gradually as they execute. In this case, we imagine that the CLM indicates what fraction of a job's processors will be available at time t.
For example, a job that has been running on 10 processors for 6
minutes might have a 
chance of completing in the next 2
minutes, releasing all ten of its processors. As an approximation of
this behavior, we predict that the job will release of its
processors within the next two minutes.
Thus we predict that the number of free processors at time t will be the sum of the processors released by each job:
Then to estimate the expected (mean) queue time we set F = n' and solve for t.
