| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 99 === Emerging heterogeneous multi-core systems provide good computing power and have
become the mainstream of computing systems. However, we notice that there is a resource contention problem in the heterogeneous multi-core systems. To solve this problem, we exploited a global resource management method to manage the heterogeneous computing resources in a system. The resource manager needs an efficient algorithm to manage the resources. We compared three algorithms in this Thesis and we found out that the min-min algorithm gives the shortest makespan among these algorithms but the average turnaround time is high. To solve this problem, we proposed a new method called MMA based on the idea of the min-min algorithm, while adding an aging technique, which is used to decrease the average turnaround time. We implemented a simulator to evaluate the performance of four algorithms, including the MMA algorithm, the Min-min algorithm, the Max-min algorithm, and the MET algorithm. Experiments show that the Min-min algorithm still has the shortest makespan, but the MMA algorithm is only left behind by about 1.5% in the first type of test cases and by the range between 0.4% and 6.6% in the second type of test cases. The MMA algorithm gives the shortest average turnaround time in all test cases and supercedes other three algorithms by 20%.
|
|---|
