DEVELOPMENT OF A PROTOTYPE APPLICATION FOR ANOMALY DETECTION BY TRACKING EVENTS IN DISTRIBUTED COMPUTER SYSTEMS ON THE EXAMPLE OF A BANKING NOTIFICATION SYSTEM

The article presents the development of a prototype application for detecting and analysing anomalies in distributed computer systems by tracking events, implemented on the example of a banking notification system. The developed prototype consists of two main modules: an anomaly detection module and a root cause analysis module. The anomaly detection module can analyse data generated by the system under surveillance through the distributed event tracking system and identify anomalous behaviour, especially those that affect the user experience. The root cause analysis module is designed to analyse the detected anomalous behaviour in the context of service dependencies and identify the service root cause of the anomalous behaviour, taking into account the effect of anomaly propagation.The paper describes in detail the architecture and principles of operation of both modules, their interaction through the Apache Kafka asynchronous communication system, and the specifics of implementing algorithms for detecting various anomalies. Particular attention is paid to detecting errors, violations of thresholds, and analysis of service response times. An innovative approach to reconstructing the structure of anomaly dependencies based on timestamps and trade identifiers is presented, allowing us to accurately determine the root causes of problems even without complete information about all spans. The prototype was developed using modern technologies and open-source frameworks, including Spring Cloud Sleuth for event tracking, Apache Kafka for asynchronous communication, and Apache Spark for data processing. The system’s architecture is designed to be scalable and add new anomaly detection algorithms easily. The study’s results demonstrate the proposed approach’s effectiveness in detecting anomalies and determining their root causes in distributed computer systems. It accurately identifies the root cause services, which can significantly speed up diagnosing and troubleshooting complex distributed systems.