Design And Implementation Of An Efficient Cloud Computing Resource Management Platform

Cloud computing is a significant computing environment providing dynamic, flexible and scalable system infrastructure in pay-as-you-go manner to the public. Server virtualization technology forms the foundation of the cloud computing with the abilities to accommodate several virtual machines on the same physical server, to regulate computing capacity of individual virtual machine (VM), and to migrate VMs across different physical servers. Hence, autonomic management of virtualized resource allocation for cloud computing has been a major challenge issue for cloud providers.In a cloud computing environment, an application can achieve load balance and workload dispersion through many executed VM entities to improve its overall performance. The goal of this research is to establish an optimized resource allocation mechanism. With the proposed scheme, we can make adjustment of resource allocation by monitoring the resource utilization, which includes the establishment and the recycling of VMs, and the resource adjustment of CPUs and RAMs. We can also predict the possibility of the resource utilization in the next time period to make some response in advance, such as increasing or decreasing VMs to guarantee the Service Level Agreement (SLA) between the service provider and the clients.In the process of resource allocation, we proposed a resource monitor mechanism for accurately and timely reflecting the variation of current resource utilization by periodically detecting the application execution time, system resource utilization ratio, and SLA. According to these conditions, the feedback module decides whether to make the immediate resource adjustment. On the other hand, the prediction module, which adopts the concept inspired from the stock market analysis, forecasts the resource utilization tendency to infer possible resource demands for the next time period. Then the arbiter module will make a suitable resource allocation policy according to the instructions offered by the feedback module and the prediction module.In addition, how to manage enough and suitable resources for applications and effectively allocate resources for physical servers is also important. To reach this goal, we proposed a stock market match approach, called the Match Maker Module, to formulate and solve the resource distribution problem between applications and the physical servers. The module would search for appropriate physical servers to produce new VMs for applications based on the available resources of physical servers and the resource allocation agreement suggested by the arbiter module. The simulation results verify the effectiveness of the proposed work.