Design And Implementation Of Edge Container Management Platform Supporting Anomaly Detection

In recent years,edge computing has been widely introduced in industrial production,processing massive data through edge devices,and realizing timely discovery of device and platform anomalies through anomaly detection algorithms.The traditional container management platform requires operators to have rich relevant knowledge,and the application threshold is relatively high;the data dimension referred to by the traditional anomaly detection algorithm is relatively single,and it is impossible to comprehensively evaluate the system operation status.In response to the above problems,this thesis designs and implements an edge container management platform that supports anomaly detection.The platform supports easy deployment of edge applications,convenient customization of edge templates,and automatic management of edge resources through the management portal,and provides anomaly detection capability.At the same time,this thesis proposes an anomaly detection algorithm based on spatial-temporal feature extraction of monitoring data.The algorithm converts the input log data into a template vector through Transformer encoder,converts the trace relationship into a grayscale image,and generates a feature vector through an attention-based convolutional neural network(CNN).The data is passed through the Bi-directional Long Short-Term Memory(Bi-LSTM)to obtain anomaly detection results.The anomaly detection algorithm integrates multiple monitoring data to ensure a comprehensive analysis of the system’s operating status;the Bi-LSTM network enhances the adaptability of anomaly detection to time-series data,thereby improving detection accuracy and reducing false alarm rate.Finally,through a series of tests,this method outperforms the baseline methods such as Deeplog and LogRobust,which proves the effectiveness of the system.This thesis first introduces the relevant background of the edge container management platform that supports anomaly detection,investigates related products and research status in the industry,and then summarizes the functional requirements of the edge container management platform that supports anomaly detection;then introduces the anomaly detection algorithm based on the spatial-temporal characteristics of platform monitoring data and its experimental results;then explained the overall system architecture design and the design of each module;then explained the system implementation.