Linux Virtual Classroom Management with Load Balancing and Computing Environment Recovery

碩士 === 國立中興大學 === 資訊科學與工程學系所 === 98 === This paper presents how the virtualization technology can be applied to build a Linux virtual classroom so that each student owns his proprietary operating environment. For virtual classroom management, even though virtualization technology has the advantages...

Full description

Bibliographic Details
Main Authors: Te-Hung Liu, 劉德鴻
Other Authors: Shang-Juh Kao
Format: Others
Language:zh-TW
Published: 2010
Online Access:http://ndltd.ncl.edu.tw/handle/65407740147919568040
Description
Summary:碩士 === 國立中興大學 === 資訊科學與工程學系所 === 98 === This paper presents how the virtualization technology can be applied to build a Linux virtual classroom so that each student owns his proprietary operating environment. For virtual classroom management, even though virtualization technology has the advantages of high availability and scalability, both problems of load imbalance among real hosts and the requirement of same computing environment as a student re-logins to the system are necessary to be resolved in an efficient way. Using the number of CPUs as the initial weight, we adopt a weighted round robin scheduling, which is built in Linux Virtual Server (LVS), to achieve the initial load balancing. And, through proposed virtual Linux Laboratory Management System (Virtual Linux Lab Management System, VLLMS), we periodically collect the CPU utilizations of real hosts. Then, by taking utilizations as new weights to re-execute LVS scheduling. We are able to accomplish the on-going load balancing. In addition, in order to allow students to restore the working environment when they re-login to the system, we make use the technology of both NFS (Network File System) and Copy on Write (CoW). Since the changed Linux environment of each student is recorded in an overlay (ovl) file, and this file is stored in a NFS shared folder, no matter where a student signs in to the classroom, the previous working environment will be brought back to the student transparently. To evaluate the feasibility, we use four Linux Kernel-based Virtual Machines (KVMs) to build an Linux virtual classroom, which is able to accommodate 20 students. In our simulation experiments, students randomly issue job requests from a pre-defined Job Pool. In every 30 seconds, we adjust the scheduling according to the new CPU utilizations. In addition to meet the load balancing, the simulation also demonstrates the successful recovery of working environment.