Dissertation > Excellent graduate degree dissertation topics show

Research of Scheduling Policies in Cloud Computing

Author: WuYiHua
Tutor: CaoJian
School: Shanghai Jiaotong University
Course: Applied Computer Technology
Keywords: The allocation of a virtual machine Demand forecast Energy consumption Self-optimization Sort model
CLC: TP3
Type: Master's thesis
Year: 2011
Downloads: 206
Quote: 0
Read: Download Dissertation

Abstract

With cloud computing technology continues to evolve and mature, the user needs tend to be diverse, the type of application tends to be complex, the resource allocation, load balancing and scheduling control and other aspects of a more specific and detailed requirements. Scheduling algorithm focuses on the traditional grid computing environment to enhance the efficiency of the implementation of the mandate, and commercial characteristics of the cloud computing environment must introduce the resources, energy consumption and other factors to create model to study the scheduling policy. At the same time, the integration of scheduling policy, with a graphical interface virtualization management platform to achieve efficient management of the cloud computing environment is very necessary. Cloud computing environment, maintaining a distinct physical configuration of the host and the normal operation of the ancillary facilities need to consume a lot of energy, in order to control the cloud computing environment operating expenses and improve its energy efficiency, this paper, based on demand forecasts virtual machine energy allocation method. First of all, due to user demand typically degeneration and meet certain seasonal model, so the use of the Holt-Winters exponential smoothing demand for subsequent cycles to predict. Secondly, according to the forecast results, the use of modified knapsack algorithm between hosts rational allocation of virtual machines. Finally, the use of self-optimization module adaptive update of the parameters in the prediction model, and to determine the appropriate forecasting cycle. The experiments show that this method can effectively reduce the the host switch operating times, thereby reducing unnecessary energy consumption in the cloud computing environment. Limited resources in order to meet the conditions the demand random and practical application environment, this article by adding the limiting factor in re-modeling, a virtual machine scheduling algorithm-based ranking model, added on the basis of the information in the original time dimension to describe demand start time, end time and duration of the first-come, first-served principle, minimizing the unallocated number of instances of the target. In this paper, application example shows the feasibility of the proposed algorithm, reasonable trigger mechanism of the algorithm by simulation. Finally, based on the discussion of the scheduling model and virtualization management platform implements a prototype system. The prototype system combines these two virtual machines scheduling algorithm, combined with the XEN virtualization technology, virtualization management platform for the user interface to the browser. Describe the system architecture of the platform through a variety of software design, module division, use case analysis, implementation details.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology
© 2012 www.DissertationTopic.Net Mobile