Àá½Ã¸¸ ±â´Ù·Á ÁÖ¼¼¿ä. ·ÎµùÁßÀÔ´Ï´Ù.
KMID : 0896219970170010119
Journal of Daegu Health College
1997 Volume.17 No. 1 p.119 ~ p.131
Task Allocation technique with minimal cost in multiprocessor system
Kang Yeong-Wook

Abstract
In multiprocessor system, most application problems can be divided into tasks for parallel execution. Scheduling these tasks can be either static or dynamic. In static scheduling, tasks are assigned to processors in compile-time. However, it is very difficult to estimate the execution time of each task before execution starts. In case that the execution times of individual tasks vary from the estimated values, optimality of the static scheduling will be lost. Dynamic task allocation can reduce the effect. But it is done during run-time, the run-time overhead becomes a critical factor. Many efforts have been made to develop low-overhead techniques to dynamically allocate tasks during job execution. With dynamic task allocation technique, it is very critical to decide which tasks to execute. This paper proposes a task selection criterion by which a task is selected from task pool for execution. It was tested and compared with previous work. It was shown that the technique with the proposed criterion has similar performance to the previous work and better than the original static schedules.
KEYWORD
FullTexts / Linksout information
Listed journal information