ADAMAS

Here is the public repository for our paper, "ADAMAS: Adaptive Domain-Aware Performance Anomaly Detection in Cloud Service Systems," which was accepted by ICSE 2025, Research Track. In this paper, we propose an adaptive and domain knowledge-aware approach to detecting performance anomalies in different services.

Overview

The following figure shows an overview of our ADAMAS framework, an adaptive performance anomaly detection framework incorporating domain knowledge.

Datasets

Due to Huawei Cloud's privacy policy, we can only release a portion of the datasets used in our study.

Requirements

Python $\geq$ 3.8 is needed. Besides, the environment can be built by: pip install -r requirements.txt

Run the code

Cd to ADAMAS/ and simply run: python ADAMAS_demo.py