Volume 38, Number 5, October 2020
|Page(s)||1030 - 1037|
|Published online||08 December 2020|
A Time Series Prediction Approach Based on Hybrid Tuning for Database Performance Indicator in AIOps
School of Computer, Northwestern Polytechnical University, Xi'an 710072, China
2 Key Laboratory of Big Data Storage and Management, Ministry of Industry and Information Technology, Xi'an 710072, China
3 Bank of Communication, Shanghai 310115, China
One of the most important applications of the intelligent operation and maintenance of a cloud database is its trend prediction of key performance indicators (KPI), such as disk use, memory use, etc. We propose a method named AutoPA4DB (Auto Prophet and ARIMA for Database) to predict the trend of the KPIs of the cloud database based on the Prophet model and the ARIMA model. Our AutoPA4DB method includes data preprocessing, model building, parameter tuning and optimization. We employ the weighted MAPE coverage to measure its accuracy and use 6 industrial datasets including 10 KPIs to compare the AutoPA4DB method with other three time-series trend prediction algorithms. The experimental results show that our AutoPA4DB method performs best in predicting monotonic variation data, e.g.disk use trend prediction. But it is unstable in predicting oscillatory variation data; for example, it is acceptable in memory use trend prediction but has poor accuracy in predicting the number of database connection trends.
云数据库智能运维中的重要应用场景之一是对监控采集的大量性能时序数据进行趋势预测。提出一种基于Prophet模型和ARIMA模型的综合调优智能趋势预测方法AutoPA4DB（auto prophet and ARIMA for database）。该方法根据数据库性能监控数据的特征，进行了原始监控数据的预处理、预测模型自动调参和模型优化。采用加权的时序预测准确性度量WMC（weighted MAPE coverage），基于多个企业级数据库实例（包含10种性能指标）进行了实验验证。实验对比了5种不同时序模型的预测效果，结果表明在单调变化模式（如磁盘使用量）的数据中，文中提出的AutoPA4DB方法时序预测准确性最高；然而在震荡模式的数据中，预测效果不太稳定，例如内存使用率趋势预测效果较好，但数据库连接数趋势预测效果不理想。
Key words: intelligent operation and maintenance / time series / prophet / ARIMA / database performance monitor
关键字 : 智能运维 / 时序数据 / Prophet模型 / ARIMA模型 / 数据库性能监控
© 2020 Journal of Northwestern Polytechnical University. All rights reserved.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.