基于深度置信网络的云应用负载预测方法  

Load Prediction Approach for Cloud Application Based on Deep Belief Networks

作　　者：马安香[1] 张长胜[1] 张斌[1] 张晓红[1] 

机构地区：[1]东北大学计算机科学与工程学院,辽宁沈阳110169

出　　处：《东北大学学报（自然科学版）》2017年第2期209-213,共5页Journal of Northeastern University(Natural Science)

基　　金：国家科技支撑计划项目(2014BAI17B00);国家自然科学基金资助项目(61572116;61572117;61502089)

摘　　要：为了准确预测云应用负载以便及时执行云应用自适应优化,从而保证云应用性能的稳定,根据云环境下应用负载预测问题的特点,提出了基于深度置信网络的云应用负载预测方法.首先给出能够有效描述负载数据的显式特征和隐式特征并定义了负载预测模型,进而给出基于深度置信网络的负载预测算法.对算法进行了分析并在真实数据集上与相关算法进行了比较,结果表明,本文提出的方法能够更加有效地解决云应用负载预测问题.To implement the adaptive optimization to ensure the performance of cloud application, it is necessary to accurately predict the load for cloud application. According to the feature of load prediction in cloud application,an approach is proposed for load prediction based on deep belief networks. Explicit and implicit features for load data are given. Load prediction model is defined. Then, the algorithm of load prediction based on deep belief networks is designed and implemented. This approach is evaluated and compared with some related load prediction algorithms,which reveals very encouraging results in terms of the prediction quality.

关 键 词：云计算 云应用 深度置信网络 负载预测 自适应优化 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]