In this study, we develop a new static scheduling scheme which integrates parallel programming environments with parallel database systems to optimize program execution. In parallel programming, a sequential program is first converted to a task graph either with programmer guidance or by a restructuring compiler. Next, a scheduling algorithm assigns the nodes of the task graph to processors. However a question arises when some tasks have to access a parallel database system. Our scheme extends static list scheduling approach to efficiently execute database accesses of parallel programs. To handle database interaction, input task graph is regenerated indicating task(s) with database interaction and these tasks are modified according to parallel database system characteristics (such as query type, database partitioning knowledge etc.). Then the proposed algorithm runs on the expanded graph by using a new heuristic which is referred to as DTF (Database Task First).
Source:
O. Dikenelli, O. Ozkasap, E. Ozkarahan, Scheduling Parallel Programs Involving Parallel Database Interactions. In V. Malyshkin (ed.),
Parallel Computing Technologies: Proceedings of the 4th International Conference,
Lect. Notes in Comp. Sci., Vol. 1277, Springer, 1997, pp. 391-393
