Synthetic Time Series for Anomaly Detection in Cloud Microservices

TL;DR

The paper tackles the difficulty of validating anomaly detection in cloud microservice environments by introducing a synthetic time-series generation framework implemented on AWS EKS. It models realistic normal load as $N(t)$ with seasonal and trend components, injects anomalies probabilistically, and collects multimodal observability data to produce labeled datasets, including two public examples. Key contributions include a complete deployment architecture, an integrated load- and fault-injection pipeline, automated anomaly labeling, and publicly available datasets to enable robust benchmarking of anomaly detection techniques. The work aims to bridge the gap between synthetic data and production realities, enabling more reliable evaluation of monitoring and anomaly detection in microservice ecosystems.

Abstract

This paper proposes a framework for time series generation built to investigate anomaly detection in cloud microservices. In the field of cloud computing, ensuring the reliability of microservices is of paramount concern and yet a remarkably challenging task. Despite the large amount of research in this area, validation of anomaly detection algorithms in realistic environments is difficult to achieve. To address this challenge, we propose a framework to mimic the complex time series patterns representative of both normal and anomalous cloud microservices behaviors. We detail the pipeline implementation that allows deployment and management of microservices as well as the theoretical approach required to generate anomalies. Two datasets generated using the proposed framework have been made publicly available through GitHub.

Figures (6)

• Figure 1: Experimental Setup Architecture
• Figure 2: Load Function and Spawn Rate
• Figure 3: User Scenarios: New Shopper User and Returning Shopper User
• Figure 4: Metrics Collection
• Figure 5: Effect of various anomalies deployed on the Front-end pod on key metrics