Cloud Systems for the Efficacy of Web Applications

• Main Authors: Chien-Te Lu, 呂建德
• Other Authors: Chiunn-Shyong Yeh
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/tg9558

博士 === 國立臺北科技大學 === 電機工程系所 === 105 === Since the invention of computers, the performance of computing systems has been an important research and practical topic. Performance attributes include, but not limit to, response time (RT), throughput, efficiency, scalability, and CPU utilization. Cloud computing, virtualization, and multicore are leading-edge technologies, which dramatically influence the development of current IT systems. In this thesis, the performance attributes are studied for virtualized Web systems running on private clouds and hybrid clouds with multicore servers. Virtual machines (VMs) for a Web application are built. Distributed stress tests are used to measure RTs, throughputs, and CPU utilization under varied combinations of virtual cores (VCs) and VM instances. Their gains, efficiencies and scalabilities are also computed and compared. For the performance of private clouds, our experimental and analytic results indicate: 1) A system can perform and scale much better by adopting multiple single-VC VMs than by single multiple-VC VM. 2) The system capacity gain is proportional to the number of VM instances run, but not proportional to the number of VCs allocated in a VM. 3) A system with more VMs or VCs has higher physical CPU utilization, but lower vCPU utilization. 4) The maximum throughput gain is less than VM or VC gain. 5) Per-core computing efficiency does not correlate to the quality of VCs or VMs employed. For the performance of hybrid clouds, our analytic results verify that hybrid clouds can reach the performance level of private clouds. They also indicate that both the inverse of RT reduction rate and throughput gain are smaller than the corresponding vCPU gain. The network latency of hybrid Clouds obviously affects CPU utilization and performance in some cases. Public cloud providers offer various deployment models with different price policies. The performance-cost relation of hybrid clouds is derived from experimental data. It offers a guideline for selecting suitable deployment model for performance requirements under a cost constraint. For an arbitrary upper bound of cost, the system using the model of less computing capacities performs better than the one using the model of more computing capacities. This thesis presents a novel method to adopt cloud technology to achieve the balance of security, performance, server sprawl, and CPU utilization. The outcomes provide valuable guidelines to select suitable instance types to use and to design proper load balancing scheme for Web systems.