@misc{17043,
  author = {Sehrish Malik and Moeen Naqvi and Leon Moonen},
  title = {Using Fault Injection to Generate Labeled Anomaly Datasets for Improving Fault Tolerance in Microservice-based Distributed Systems},
  abstract = {The complex nature of the microservices infrastructure, coupled with the shared and dynamic environment of the cloud, poses challenges for fault management, making it more challenging than other types of systems. This paper proposes a novel approach, SynFI, for generating automated synthetic labeled datasets that address the challenges of effectively monitoring highly distributed microservice systems and making them robust and faulttolerant. SynFI builds on chaos engineering principles, systematically exposing the system to faults and recording observability data under normal and faulty conditions using monitoring tools. Our approach extracts the fault labels and data metrics from the collected data and maps them onto a synthetic labeled dataset that includes accurate fault labels. We analyze the collected data through data quality analysis and test the prediction accuracy for anomaly detection of the dataset using six popular classification algorithms. Overall, the paper contributes to developing self-adaptive and self-healing microservices-based systems, leading to improvements in administration, allocation of microservices across servers, and quality attributes like reliability, scalability, and resilience.},
  year = {2023},
  publisher = {Simula Research Laboratory},
  address = {Oslo, Norway},
}