On the twin measure and queueing systems predictability

Two identical jobs with deterministically identical processing times arrive at a web server simultaneously (Twins), but leave the system thirty seconds apart. Is the service predictable? Is their sojourn time predictable? This issue arises in modern day networking systems such as call centers and web servers as well as in other queueing systems. We propose a novel measure based on the principle that in a predictable system, “twin” jobs should not depart the system very far apart. We analyze this measure for a number of common scheduling policies and compare the results. We compare the results to those of other predictability approaches proposed recently and discuss the measure’s usefulness.