Article Preview
Top1. Introduction
Computers are an indispensable tool in modern offices and have become an important foundation for the development of all walks of life. With the passage of time and the rapid development of technology, traditional computer office systems have gradually exposed many shortcomings. When the computer system is large in scale, it often faces many problems such as heavy maintenance tasks, difficult equipment management, and high operation and maintenance costs, and the remote deployment and management of branch offices becomes more difficult. In order to solve the dilemma faced by traditional IT systems, this article uses a combination of cloud computing and traditional desktop virtualization technology to replace traditional IT systems.
Virtual machine redistribution involves the rescheduling or matching of virtual machines and physical machines to meet the changes in virtual machine resource requirements and provide scalability and reliability Guerrero et. al. (2018). Ferreto et al. Minimize the number of virtual machines that need to be migrated to improve the state of the host by avoiding the reallocation of virtual machines with stable resource requirements Han et. al. (2015). This method can reduce the migration of virtual machines to some extent, but some aspects such as resource utilization and power are not fully considered. Ghribi et al. Proposed an algorithm to control the number of virtual machines that need to be migrated and minimize power consumption Sheng, et. al. (2015). Maguluri et al. Used a random model to monitor the resource load on each host and used it as a reference for new assignments at each time interval. In order to maximize the output of the entire system without affecting the performance of the entire system, the operation of virtual machine redistribution needs to be performed periodically Zhong, et. al. (2016). Feller et al. Proposed the use of ant colony-based algorithm to complete the dynamic matching of virtual machines Okada, et. al. (1987). Utilize the method of host integration to reduce power consumption by increasing the resource utilization of some hosts and reducing the number of physical hosts used.
This article uses the ARIMA time series prediction model to predict the utilization of host resources at the next moment, and then determines whether to trigger the migration based on a preset double threshold. By fully studying the data of host resource utilization, the data's volatility and trend are obtained. In the virtual machine selection strategy, the overload trigger algorithm is designed to fully consider the migration cost and time of the virtual machine, and use the memory and CPU factors to build a model, and use this model to select the appropriate virtual machine to be migrated. In the process of virtual machine dynamic migration, the evaluation functions of SLA, resource balance and power loss were designed according to the optimization goals proposed in this paper Lashgar, et. al. (2016). Using these evaluation functions and the BFD binning algorithm, a multi-target dynamic migration algorithm for virtual machines is proposed. After researching related desktop virtualization solutions, the architecture of the desktop virtualization system in this paper is designed and implemented and deployed under OpenStack Cloud Desktop virtualization system Dube, et. al. (2017). During the design of the desktop virtualization system architecture, the decoupling between systems and the security of data were considered, and the system was modularized to improve the scalability of the system.
In this paper, a dual-threshold trigger strategy is used. Below the lower threshold, a low-load trigger algorithm is used, and above the upper threshold, an overload trigger algorithm is used to obtain the virtual machine to be migrated. Using evaluation function and BFD boxing algorithm, a multi-target dynamic migration algorithm for virtual machines is proposed, and the effectiveness of the dynamic management strategy for virtual machines in this paper is verified by experiments. Finally, the availability of the system and the validity of related theoretical results are verified through experiments.