Complexity Results for Scheduling Tasks in Fixed Intervals on Two Types of Machines

Suppose that n independent tasks are to be scheduled without preemption on an unlimited number of parallel machines of two types: inexpensive slow machines and expensive fast machines. Each task requires a given processing time on a slow machine or a given smaller processing time on a fast machine. We make two different feasibility assumptions: (a) each task has a specified processing interval, the length of which is equal to the processing time on a slow machine; (b) each task has a specified starting time. For either problem type, we wish to find a feasible schedule of minimum total machine cost. It is shown that both problems are NP-hard in the strong sense. These results are complemented by polynomial algorithms for some special cases.