When jobs have to be processed on a set of identical parallel machines so as to minimize the makespan of the schedule, list scheduling rules form a popular class of heuristics. The order in which jobs appear on the list is assumed here to be determined by the relative size of their processing times; well known special cases are the LPT rule and the SPT rule, in which the jobs are ordered according to non-increasing and non-decreasing processing time respectively.
When all processing times are exactly known, a given list scheduling rule will generate a unique assignment of jobs to machines. However, when there exists a priori uncertainty with respect to one of the processing times, then there will be, in general, several possibilities for the assignment that will be generated once that processing time is known. That number of possible assignments may be viewed as a measure of the sensitivity of the list scheduling rule that is applied.
We derive bounds on the maximum number of possible assignments for several list scheduling heuristics, and we also study the makespan associated with these assignments. In this way we obtain analytical support for the intuitively plausible notion that the sensitivity of a list scheduling rule increases with the quality of the schedule produced.
Key words: sensitivity analysis, list scheduling, heuristics, robustness, scheduling.