GIED

This repo contains the simulation environment dataset and source code showpieces for the paper "Graph based Incident Extraction and Diagnosis in Large-Scale Online Systems" (ASE'22).

Project Structure

• ./data contains the simulation environment dataset used in the paper. The real-world dataset from the company cannot be provided here yet due to the confidentiality policy of the company.
• ./showpieces contains ipython notebooks which run some code pieces of GIED to show how each step is performed. Their order is as follow:
  • anomaly_detection_and_issue_extraction.ipynb contains code pieces for KPI anomaly detection and issue extraction.
  • data_labelling.ipynb contains code pieces for data labelling using fault injection records.
  • feature_engineering.ipynb contains code pieces for feature engineering.
  • SpatioDevNetPackage contains the implemented graph neural networks based model.
  • incident_detection.ipynb contains code pieces for the graph neural networks based model training and testing for incident detection.
  • incident_diagnosis.ipynb contains code pieces for the root cause service localization.