The problem of scheduling a collection of different jobs on identical parallel machines is investigated. For each job a number of unit processing time tasks, a given deadline and an upper bound on inventory are known. The optimality criterion is changeover cost which occur if different types of jobs are processed in sequence. Only unit cost are to be considered. For two job types, an efficient algorithm is presented considering only tight schedules; if more job types have to be considered the problem becomes NP-hard.