Performance Enhancement of Backfilling Scheduling Methods Using Genetic Algorithm

• Main Authors: Chia-Chun Chiang, 江嘉峻
• Other Authors: 林偉
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/68590069877288768281

碩士 === 國立中興大學 === 資訊網路多媒體研究所 === 100 === As the underlying network equipments become complete and mature, cloud computing has become the future mainstream of development toward the goal of green IT, energy saving and carbon reduction. A variety of research efforts have been devoted to the performance improvement of the cloud networks. Major issues under study in the cloud environment include task scheduling, virtual machine management, and energy consumption. In this thesis, we focus on scheduling tasks in the cloud network environment. We propose a backfilling-based scheduling method for achieving efficient and quality task scheduling. Our method features the use of the genetic algorithm with balanced spiral for scheduling the tasks. The main idea behind our research effort is to rearrange a sequence of tasks with the aid of the genetic algorithm with balanced spiral. We aim at improving the performance of the backfilling scheduler so that the completion time and the average waiting time of a given task sequence are minimized. We choose the cloud network simulator – CloudSim - as an experimental platform for verification. The simulation results show that our proposed method considerably improves utilization, when compared with the basic backfill algorithm. This leads to a significant reduction in the completion time and the average waiting time for a given task sequence, especially when these tasks have drastically different lengths. For example, in a case of 5000 tasks with various lengths, the proposed method achieves an 8% reduction in completion time. Through the simulation, we also find two major factors that can influence the task completion time: fixed or variable task length, and ascending or descending task sequence.