ScholarWorks@숙명여자대학교

초록

Cloud computing technology enables scientists to dynamically expand their environments for scientific experiments. However, to maximize performance and satisfy user requirements it is difficult to quickly provide hybrid resources suitable to application characteristics. In this paper, we design a resource provisioning model based on application characteristic profiles and job history analysis in hybrid computing infrastructure consisting of cluster and cloud environments. In addition to the multi-layer perceptron machine learning method, error backpropagation technique is used to analyze job history to re-learn the error of the output value. Also, we propose an adaptive resource provisioning method for horizontal/vertical scaling of VMs in accordance with the state of the system. We experiment CPU-intensive applications according to the proposed model and algorithms, in a hybrid infrastructure. The experimental results show that using the proposed method, we satisfy user-specified SLA (cost and execution time) and improve the efficiency of resource usage. © 2016 IEEE.

키워드