《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (5): 1508-1515.DOI: 10.11772/j.issn.1001-9081.2021030393

• 先进计算 • 上一篇    下一篇

基于长-短时序特征融合的资源负载预测模型

王艺霏1, 于雷2,3, 滕飞1(), 宋佳玉1, 袁玥1   

  1. 1.西南交通大学 信息科学与技术学院, 成都 610000
    2.北京航空航天大学 中法工程师学院, 北京 100000
    3.北京航空航天大学 杭州创新研究院(余杭), 杭州 310000
  • 收稿日期:2021-03-16 修回日期:2021-06-08 接受日期:2021-06-11 发布日期:2022-06-11 出版日期:2022-05-10
  • 通讯作者: 滕飞
  • 作者简介:王艺霏(1996—),女,山西吕梁人,硕士研究生,主要研究方向:云计算、大数据挖掘
    于雷(1972—),男,山东淄博人,副教授,博士,主要研究方向:大数据计算中心能耗优化与仿真、基于深度学习的图像处理、自然语言处理与分类
    滕飞(1984—),女,山东淄博人,副教授,博士,CCF会员,主要研究方向:云计算、医疗信息、工业大数据挖掘 fteng@swjtu.edu.cn
    宋佳玉(1995—),女,四川成都人,硕士研究生,主要研究方向:大数据挖掘、时间序列预测
    袁玥(1997—),女,四川泸州人,硕士研究生,主要研究方向:大数据挖掘、深度学习。
  • 基金资助:
    四川省科技项目(2019YJ0214);北京市自然科学基金资助项目(4192030)

Resource load prediction model based on long-short time series feature fusion

Yifei WANG1, Lei YU2,3, Fei TENG1(), Jiayu SONG1, Yue YUAN1   

  1. 1.School of Information Sciences and Technology,Southwest Jiaotong University,Chengdu Sichuan 610000,China
    2.Sino? french Engineer School,Beihang University,Beijing 100000,China
    3.Beihang Hangzhou Institute for Innovation at Yuhang,Hangzhou Zhejiang 310000,China
  • Received:2021-03-16 Revised:2021-06-08 Accepted:2021-06-11 Online:2022-06-11 Published:2022-05-10
  • Contact: Fei TENG
  • About author:WANG Yifei, born in 1996, M. S. candidate. Her researchinterests include cloud computing,big data mining.
    YU Lei, born in 1972,Ph. D.,associate professor. His researchinterests include energy consumption optimization and simulation of big data computing center,image processing based on deep learning,natural language processing and classification.
    TENG Fei, born in 1984,Ph. D.,associate professor. Her researchinterests include cloud computing,medical informatics,industrial big data mining.
    SONG Jiayu, born in 1995, M. S. candidate. Her researchinterests include big data mining,time series prediction.
    YUAN Yue, born in 1997,M. S. candidate. Her research interestsinclude big data mining,deep learning.
  • Supported by:
    Science and Technology Project of Sichuan Province(2019YJ0214);Natural Science Foundation of Beijing(4192030)

摘要:

高准确率的资源负载预测能够为实时任务调度提供依据,从而降低能源消耗。但是,针对资源负载的时间序列的预测模型,大多是通过提取时间序列的长时序依赖特性来进行短期或者长期预测,忽略了时间序列中的短时序依赖特性。为了更好地对资源负载进行长期预测,提出了一种基于长-短时序特征融合的边缘计算资源负载预测模型。首先,利用格拉姆角场(GAF)将时间序列转变为图像格式数据,以便利用卷积神经网络(CNN)来提取特征;然后,通过卷积神经网络提取空间特征和短期数据的特征,用长短期记忆(LSTM)网络来提取时间序列的长时序依赖特征;最后,将所提取的长、短时序依赖特征通过双通道进行融合,从而实现长期资源负载预测。实验结果表明,所提出的模型在阿里云集群跟踪数据集CPU资源负载预测中的平均绝对误差(MAE)为3.823,均方根误差(RMSE)为5.274,拟合度(R2)为0.815 8,相较于单通道的CNN和LSTM模型、双通道CNN+LSTM和ConvLSTM+LSTM模型,以及资源负载预测模型LSTM-ED和XGBoost,所提模型的预测准确率更高。

关键词: 资源负载预测, 卷积神经网络, 长短期记忆网络, 格拉姆角场, 双通道, 时间序列预测

Abstract:

Resource load prediction with high accuracy can provide a basis for real-time task scheduling, thus reducing energy consumption. However, most prediction models for time series of resource load make short-term or long-term prediction by extracting the long-time series dependence characteristics of time series and neglecting the short-time series dependence characteristics of time series. In order to make a better long-term prediction of resource load, a new edge computing resource load prediction model based on long-short time series feature fusion was proposed. Firstly, the Gram Angle Field (GAF) was used to transform time series into image format data, so as to extract features by Convolutional Neural Network (CNN). Then, the CNN was used to extract spatial features and short-term data features, the Long Short-Term Memory (LSTM) network was used to extract the long-term time series dependent features of time series. Finally, the extracted long-term and short-term time series dependent features were fused through dual-channel to realize long-term resource load prediction. Experimental results show that, the Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and R-squared(R2) of the proposed model for CPU resource load prediction in Alibaba cloud clustering tracking dataset are 3.823, 5.274, and 0.815 8 respectively. Compared with the single-channel CNN and LSTM models, dual-channel CNN+LSTM and ConvLSTM+LSTM models, and resource load prediction models such as LSTM Encoder-Decoder (LSTM-ED) and XGBoost, the proposed model can provide higher prediction accuracy.

Key words: resource load prediction, Convolution Neural Network (CNN), Long Short-Term Memory (LSTM) network, Gram Angle Field (GAF), dual-channel, time series prediction

中图分类号: