ADARMA Auto-Detection and Auto-Remediation of Microservice Anomalies by Leveraging Large Language Models.

In microservice architecture, anomalies can cause slow response times or poor user experience if not detected early. Manual detection can be time-consuming and error-prone, making real-time anomaly detection necessary. By implementing runtime performance anomaly detection models, microservice systems can become more stable and reliable. However, anomaly detection alone is not enough, and complementary auto-remediation techniques are required to automatically detect and fix issues. Auto-remediation techniques can optimize resource allocation, tune code, or trigger automatic recovery mechanisms. The combination of anomaly detection and auto-remediation reduces downtime and enhances system reliability, resulting in increased productivity and customer satisfaction, which, in turn, drives higher revenue. Prior works have overlooked auto-remediation. In this work in progress paper, we propose a pipeline for automatic anomaly detection and remediation based on Large Language Models (LLMs).