FCA-based Virtual Cluster Deployment and Migration

• Main Authors: CHIA-YU CHU, 朱家玉
• Other Authors: 高勝助
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/79621364220321933609

碩士 === 中興大學 === 資訊科學系所 === 95 === To provide virtualization directly from CPU is currently a burgeoning issue in the computer science. As a consequence, Intel and AMD, the two major CPU manufacturers,have developed the Intel Virtualization Technology and AMD Secure Virtual Machine respectively. Furthermore, many virtual machine implementations have been reported since the concept of virtual machine was proposed in 1960s, such as the VW of IBM and WMware of VMware. As of its features of highly security and scalability with virtualization, virtual machines are useful for system simulation in the early days and popular for the currently widely adopted network applications, such as the establishment of sandboxes, honeypots, and all kinds of Internet servers. Due to the popularity of applications using virtual machines, in this study, we focus on the architecture of clustered environment using virtual machines and present an efficient management system, especially in the area of automatic deployment and dispatch. As Linux Virtual Server (LVS) is one of the widely used Linux cluster management software, it is capable of providing network services with high practicality and expandability for the network cluster composed of physical servers. We installed User-Mode Linux (UML) virtual machines in several physical machines to provide various network services, such as web server and ftp server. Using virtual machines for network services, the service reliability can be significantly enhanced since the failure of a virtual machine couldn’t bring down other services in the same physical machine. In addition, the failed UML machine can be easily resumed without re-starting the physical machine. In our study, a LVS cluster environment is built with the assistance of using formal concept analysis (FCA) module. Several system performance attributes are extracted, such as cpu loading, memory usage, and free disk space, from each real server. Then, we manipulate the collection and apply the partial ordering of FCA for optimal virtual machine deployment as well as the relocation due to the balancing effect. At the end, we propose an effective Linux cluster management system for the cluster environment of UML-based virtual machines. Keywords: virtual machine, cluster environment, formal concept analysis, user-model Linux, Linux virtual server