文章基本信息

标题：A Fusion Model for CPU Load Prediction in Cloud Computing
本地全文：下载
作者：Xu, Dayu ; Yang, Shanlin ; Luo, He 等
期刊名称：Journal of Networks
印刷版ISSN：1796-2056
出版年度：2013
卷号：8
期号：11
页码：2506-2511
DOI：10.4304/jnw.8.11.2506-2511
语种：English
出版社：Academy Publisher
摘要：Load prediction plays a key role in cost-optimal resource allocation and datacenter energy saving. In this paper, we use real-world traces from Cloud platform and propose a fusion model to forecast the future CPU loads. First, long CPU load time series data are divided into short sequences with same length from the historical data on the basis of cloud control cycle. Then we use kernel fuzzy c-means clustering algorithm to put the subsequences into different clusters. For each cluster, with current load sequence, a genetic algorithm optimized wavelet Elman neural network prediction model is exploited to predict the CPU load in next time interval. Finally, we obtain the optimal cloud computing CPU load prediction results from the cluster and its corresponding predictor with minimum forecasting error. Experimental results show that our algorithm performs better than other models reported in previous works.
关键词：Cloud Computing;Load Prediction;Fuzzy Clustering;Neural Network