Using Linear Programming to Solve Reliability-Constrained Task Scheduling in Computer Clusters

• Main Authors: Yu-ChangTai, 戴毓璋
• Other Authors: Ci-Ping Chu
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/84933250090521038189

碩士 === 國立成功大學 === 資訊工程學系 === 102 === In parallel computing environment, the reliability is an important issue in the scheduling problem. Optimizing both finish time and reliability in task scheduling problem is a NP-complete problem. In order to increase the reliability of task scheduling, duplicating task is a very common technique. In this thesis, we propose a method of linear programming to find the most suitable duplication number of each task as possible, such that there is fastest finish time in the scheduling. Otherwise, the reliability must meet the requirement specified by users or systems. In this study, the reliability evaluation considers both processor and communication link. The reliability is a probability representing successful implementation of an application in computer clusters. Finally, the experimental results show that the presented algorithm outperforms other ones proposed.