Ok Ki Sang, Park Joon Seok, Lee won Joo, Jeon Chang Ho

The Transactions of the Korea Information Processing Society (1994 ~ 2000), Vol. 6, No. 6, pp. 1502-1510, Jun. 1999
10.3745/KIPSTE.1999.6.6.1502,   PDF Download:

Abstract

In adaptive job allocation method for multiprocessor systems a job is folded, or split in two halves, to fit for an available subcube in order to reduce the waiting time of jobs. In this method, however, since a job is folded whenever a subcube with the proper size is not found, the prolonged execution time caused by job split may override the savings in waiting time, in which case the total execution cost of jobs may be increased. In this paper, an improved adaptive job allocation algorithm, called Estimate-fold allocation, is presented and evaluated. The proposed algorithm estimates the costs and takes the better of two alternatives; folding right away and waiting until a bigger subcube becomes available. The average total job execution cost of our algorithm is calculated and compared to those of the conventional adaptive, buddy, and gray-mode algorithms through simulations. The results shows that our proposed algorithm performs better than others.

Statistics
Cite this article