Loading...

A Hybrid Method Based on Scheduling and Load Balancing to Reduce the Power Consumption in Cloud Networks

Tarahomi, Mehran | 2019

465 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: English
  3. Document No: 52477 (52)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Izadi, Mohammad
  7. Abstract:
  8. In recent years, cloud computing has provided a well-known platform to provide on-the-fly and cost-efficient services on local and world area networks. As a result, the number of cloud data centers is increasing due to the increasing number of customers. For this reason, the energy consumption of numerous cloud data centers is one of the biggest bottlenecks for cloud providers. In general, various hardware and software-based solutions can be used to optimize power consumption at the same time of guaranty service level agreements. One of the most popular of these software-based solutions is dynamic allocation of resources to virtual machines (VM), i.e. a sandboxing technique responsible for executing cloud customer requests. In this dissertation, three VM allocation strategies have been proposed to optimize the power consumption in the cloud infrastructure. In our first endeavor, we propose a novel prediction-based resource allocation algorithm. The presented ensemble prediction algorithm predicts the resources load based on outlier detection in time slot data of the resources and uses an algorithm for model selection. In this work, we consider components in module, rack, and host levels and each level is individually controlled in a hierarchical manner from module (top) to host (bottom) level. Experimental results indicate that the proposed solution reduces the power consumption; moreover, the service level agreement violation (SLAV) is improved. In the second work, a micro-genetic VM allocation algorithm to minimize number of available hosts is proposed. To do so, VMs and hosts are modeled with a chromosome representation. The proposed VM allocation algorithm shows a robust performance and outperforms classic genetic algorithm in different situations. In our last practice, a novel fuzzy logic based algorithm is applied to find minimum required data centers based on the users-demanded VMs. Furthermore, an improved version of dynamic voltage frequency system (DVFS) algorithm and a new VM allocation for load balancing are provided. The proposed method outperforms all baseline algorithms based minimum data centers, efficient task scheduling and VM load balancing provided. All experiments reported in this thesis have been done using CloudSim toolkit, a widely used and standard simulation tool for cloud environment
  9. Keywords:
  10. Cloud Computing ; Virtualization ; Energy Productivity ; Virtual Machine Allocation ; Dynamic Consolidation ; Load Balancing ; Power Reduction

 Digital Object List

 Bookmark

No TOC