| Summary: | 碩士 === 國立高雄第一科技大學 === 資訊管理研究所 === 102 === Traditionally, an enterprise needs a lot of hardware resources to run various information systems for daily operations, like server farms and the PC clients. However, most of the hardware resources are inefficiently utilized or even idle in daily business operation under the situation that each host is usually responsible for serving only a specific computing task. Moreover, the air-conditioning equipment, uninterruptible power systems and other equipment, which are proportional to the size of the host population, will also be gradually installed in an imperceptible manner. Large enterprises may be able to afford this, but for most of the SMEs in Taiwan, this is usually supposed to be extravagant or unaffordable due to their limited resources.
For economical reasons, the SMEs in Taiwan therefore often “squeeze” a lot of planned computing services on the implementation of a single host to avoid unnecessary installations. But, this means that the administrator and maintenance group should keep striving to ward off any inevitable problem caused from hardware fails or network attacks. Therefore, to promote the competitiveness of contemporary enterprises, it is necessary to make large enterprises reduce the host installations cost-effectively, and pave a way for SMEs to efficiently manage their information systems.
As cloud computing is evolving, we intend to utilize the multi-core and multi-threading CPU hardware support, together with the virtualization technology to provide a solution for such dilemma. Under such computing environments, several virtualization systems running different services can be configured to share a common resource pool. This is also called a Scale-In solution in IT industry. In this thesis, we will use a variety of Open Source software running on Oracle VM Virtual Box to alleviate the above problem arisen in a real SME in Taiwan. We will explain how to virtualize the host resources, employ Web Services and use Apache/Jboss application technology to construct a testbed for easing IT personnel to effectively manage the same service systems without interruption. By conducting some experimental performance evaluation, we conclude that our approach and the system configuration can be effectively considered as a valuable reference for similar scale SMEs. The stress test based on the measurement tools jmeter shows that about 50% of the samples in the reaction time can be improved by 1.36%, error rate can be reduced from 1.75% to 0%, and the amount of request can be decreased from 9.9 to 3.4 per second, which are all satisfiable as expected.
|
|---|
