基于ECA-TCN的数据中心磁盘故障预测

PDF下载HTML

Title:

Disk failure prediction in data centers based on ECA-TCN

作者:

张铭泉^1,2, 王宝兴^1,2

1. 华北电力大学 控制与计算机工程学院, 河北 保定 071003;
2. 华北电力大学 复杂能源系统智能计算教育部工程研究中心, 河北 保定 071003

Author(s):

ZHANG Mingquan^1,2, WANG Baoxing^1,2

1. School of Control and Computer Engineering, North China Electric Power University, Baoding 071003, China;
2. Engineering Research Center of Intelligent Computing for Complex Energy Systems Ministry of Education, North China Electric Power University

关键词:

磁盘故障预测; 长短时记忆网络; 循环神经网络; 扩张卷积; 高效通道注意力机制; 神经网络模型; 时间序列预测; 深度学习优化

Keywords:

disk failure prediction; long short-term memory network; recurrent neural network; dilated convolution; efficient channel attention mechanism; neural network model; time series prediction; deep learning optimization

分类号:

TP391.4

DOI:

10.11992/tis.202310043

摘要:

随着数据中心规模的不断扩大，磁盘故障对数据中心的运行稳定性产生越来越大的影响。当前预测方法在面对大规模、高维度和长序列的磁盘运行数据时仍存在不足。本文提出了一种高效通道注意力时间卷积网络(efficient channel attention-temporal convolutional network, ECA-TCN)模型，通过结合传统卷积神经网络一维卷积的优势，融入扩张卷积和残差结构，并引入注意力机制，该模型能够提高磁盘故障预测的准确性和稳定性。在实验中，将ECA-TCN模型与其他经典深度学习方法进行了比较，实验结果表明，ECA-TCN模型在磁盘故障预测任务上具有较高的准确性和稳定性。

Abstract:

With the continuous expansion of the scale of the data center, disk failure has an increasing impact on the stability of the data center. Current prediction methods still have shortcomings in the face of large-scale, high-dimensional and long sequence of disk running data. This paper proposes an efficient channel attention-temporal convolutional network (ECA-TCN) model. By combining the advantages of one-dimensional convolution of traditional convolutional neural network, integrating dilated convolution and residual structure, and introducing attention mechanism, the model can improve the accuracy and stability of disk failure prediction. In the experiment, the ECA-TCN model is compared with other classical deep learning methods. The experimental results show that the ECA-TCN model has high accuracy and stability in the disk failure prediction task.