Suppose instances of the same or different workstreams share physical CPUs serially across time in the proposed configuration. How should the SLT be calculated for those instances?
As the general statement, SLT cannot be counted if a workstream instance cannot be run. Therefore, just as the throughput time for an instance blocked from starting due to physical use of the CPUs is started from the time of submission, the throughput time for an instance blocked from restarting is measured until the instance may be restarted. For example, suppose we consider a physical block of CPU resources supporting two workstream instances A and B (A and B may or may not be from the same workstream). According to the throughput definition, A and B must be submitted to the run queue at the same time. Suppose A starts. Thus B must wait for A to complete. Upon the completion of A, B starts. Therefore, the time for B is the queue wait time plus the run time for B because B could not run until A completed. Similarly, the throughput time for A is the time that A took to run plus the time B takes to run since A could not restart until B completes.
