In this paper we examine a new scheduling problem that shows up in the construction of a parallel digital library. This library performs a large amount of operations on a huge amount of stored data items. As each single operation is sequential, the data items and the sequential operation have to "meet" on one processor. Thus, a scheduling and dynamic mapping problem has to be solved to optimize the overall throughput of the server system.We present a theoretical model of this new scheduling problem and analyze heuristics for it. These heuristics use greedy strategies and local search principles; their performance is valued comparing the results with a lower bound which bases on a lp-relaxation. We have done a large amount of experiments due to get knowledge of the behavior of the heuristics.