应用n-LSTM的云平台任务CPU负载预测方法  

Cloud Platform Task CPU Load Prediction Method Using n-LSTM

作　　者：曹振 邓莉[1,2] 谢同磊 梁晨君 CAO Zhen;DENG Li;XIE Tonglei;LIANG Chenjun(College of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065,China;Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System,Wuhan 430065,China)

机构地区：[1]武汉科技大学计算机科学与技术学院,武汉430065 [2]智能信息处理与实时工业系统湖北省重点实验室,武汉430065

出　　处：《小型微型计算机系统》2024年第1期75-83,共9页Journal of Chinese Computer Systems

基　　金：新一代信息技术创新项目(2020ITA01005)资助.

摘　　要：云平台任务的CPU负载预测有助于云平台资源的优化配置,以改善资源利用率.它是有效管理云资源的重要手段.为提高任务CPU负载预测精度,本文主要做了以下工作:1)利用热度图提取用于进行CPU负载预测的资源使用特征;2)设计并实现了一种基于n-LSTM的云平台任务的CPU负载预测方法DPFE-n-LSTM;3)分别在阿里云平台数据集和Google云平台数据集上进行了实验,结果表明,相对于目前已经提出的CPU负载预测模型BP、LSTM和CNN-LSTM,DPFE-n-LSTM方法具有更好的预测性能.CPU load prediction of cloud tasks helps to optimize the allocation of cloud resources and also improves resource utilization.It is very important for effective resource management in the cloud.In order to improve the accuracy of CPU load prediction for tasks,this paper mainly did the following work;1)Use heat map to extract resource usage characteristics for CPU load prediction;2)A CPU load prediction method for cloud tasks named DPFE-n-LSTM is designed and implemented;3)Experiments were carried out both on Alibaba Cloud trace and Google Cloud trace.The results show that compared to other prediction models,such as BP,LSTM and CNN-LSTM,that have been proposed so far,DPFE-n-LSTM has better forecast performance.

关 键 词：特征选择 CPU负载 n-LSTM 时间序列 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]