LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep Learning

Microservice architecture is a high-cohesion and low-coupling software architecture. Its core idea is to split the application into a set of microservices with a single function and independent deployment. Due to their complexity and large scale, microservice systems are typically fragile and failur...

Full description

Saved in:
Bibliographic Details
Main Authors: Xu Liu, Yuewen Liu, Miaomiao Wei, Peng Xu
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10720008/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850114108708356096
author Xu Liu
Yuewen Liu
Miaomiao Wei
Peng Xu
author_facet Xu Liu
Yuewen Liu
Miaomiao Wei
Peng Xu
author_sort Xu Liu
collection DOAJ
description Microservice architecture is a high-cohesion and low-coupling software architecture. Its core idea is to split the application into a set of microservices with a single function and independent deployment. Due to their complexity and large scale, microservice systems are typically fragile and failures are inevitable. Therefore, there is an urgent need for fast and accurate anomaly detection capabilities. However, the existing microservice anomaly detection methods do not pay attention to the multi-source data of the microservice system and thus have low accuracy. To address this limitation, we propose a Log-Metric Combined Microservice Anomaly Detection approach through Graph-based Deep Learning (termed as LMGD). First, we propose a time-aware LSTM prediction neural network to improve the accuracy of service dependency mining. Secondly, based on the service dependency graph, we propose an anomaly detection method based on log-metric fusion, which can more accurately describe the running status of microservices, thereby improving the accuracy of anomaly detection. The experimental outcomes demonstrate that compared with other state-of-the-art methods, our method improves recall and F1-score by 2.63% and 1.05%.
format Article
id doaj-art-451b7f1f858d4efaa1ba63459333982e
institution OA Journals
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-451b7f1f858d4efaa1ba63459333982e2025-08-20T02:36:59ZengIEEEIEEE Access2169-35362024-01-011218651018651910.1109/ACCESS.2024.348167610720008LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep LearningXu Liu0Yuewen Liu1https://orcid.org/0009-0005-7964-3432Miaomiao Wei2Peng Xu3https://orcid.org/0000-0003-3635-2342China Academy of Industrial Internet, Beijing, ChinaState Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, ChinaState Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, ChinaState Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, ChinaMicroservice architecture is a high-cohesion and low-coupling software architecture. Its core idea is to split the application into a set of microservices with a single function and independent deployment. Due to their complexity and large scale, microservice systems are typically fragile and failures are inevitable. Therefore, there is an urgent need for fast and accurate anomaly detection capabilities. However, the existing microservice anomaly detection methods do not pay attention to the multi-source data of the microservice system and thus have low accuracy. To address this limitation, we propose a Log-Metric Combined Microservice Anomaly Detection approach through Graph-based Deep Learning (termed as LMGD). First, we propose a time-aware LSTM prediction neural network to improve the accuracy of service dependency mining. Secondly, based on the service dependency graph, we propose an anomaly detection method based on log-metric fusion, which can more accurately describe the running status of microservices, thereby improving the accuracy of anomaly detection. The experimental outcomes demonstrate that compared with other state-of-the-art methods, our method improves recall and F1-score by 2.63% and 1.05%.https://ieeexplore.ieee.org/document/10720008/Anomaly detectionlogmetricsdistributed systemsGCNGAT
spellingShingle Xu Liu
Yuewen Liu
Miaomiao Wei
Peng Xu
LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep Learning
IEEE Access
Anomaly detection
log
metrics
distributed systems
GCN
GAT
title LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep Learning
title_full LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep Learning
title_fullStr LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep Learning
title_full_unstemmed LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep Learning
title_short LMGD: Log-Metric Combined Microservice Anomaly Detection Through Graph-Based Deep Learning
title_sort lmgd log metric combined microservice anomaly detection through graph based deep learning
topic Anomaly detection
log
metrics
distributed systems
GCN
GAT
url https://ieeexplore.ieee.org/document/10720008/
work_keys_str_mv AT xuliu lmgdlogmetriccombinedmicroserviceanomalydetectionthroughgraphbaseddeeplearning
AT yuewenliu lmgdlogmetriccombinedmicroserviceanomalydetectionthroughgraphbaseddeeplearning
AT miaomiaowei lmgdlogmetriccombinedmicroserviceanomalydetectionthroughgraphbaseddeeplearning
AT pengxu lmgdlogmetriccombinedmicroserviceanomalydetectionthroughgraphbaseddeeplearning