摘要 |
<p>Disclosed is a scheduling device for improving system performance and usage efficiency of resources when time-sharing a parallel data processing system by a plurality of dataflow programs. A scheduling device comprises a program analysis unit that refers to first and second dataflow programs and generates first and second DAGs representing both dataflow programs, and first and second processing execution order information units representing execution order of processing corresponding to both DAG nodes; a processing assignment unit for assigning processing corresponding to nodes of the first and second DAGs to a plurality of data processing devices on the basis of the first and second processing execution order information units; and a data assignment unit for referring to the first and second DAGs and the first and second processing execution order information units, and exchanging data used in the processing corresponding to the nodes of the first and second DAGs between a storage unit of a storage device and storage units provided in each of the plurality of data processing devices.</p> |