Replication package

This repo contains the implementation details for the paper "Cross-System Categorization of Abnormal Traces in Microservice-Based Systems via Meta-Learning"

File structure and description

• OnlineBoutiqueFaultList: a list of base fault categories and novel fault categories in our study
• TrainticketFaultList: a list of base fault categories and novel fault categories in our study
• models:
  • MAML.py: the implementation of the MAML algorithm. The MAML class manages the meta-training and meta-testing phases. It trains the base model in Learner.py. We considered this repo MAML-Pytorch in our implementation.
  • Learner.py: defines the base model that do abnormal trace classification
  • AttenAE.py: the implementation of AttenAE
  • DatasetFusion.py: load unlabled traces into the Dataset
  • AttenAE_train.py: train AttenAE using unlabled traces for each MSS
  • DatasetMix.py: Dataset for cross-system contexts
  • DatasetTT.py: Dataset only for Trainticket-Trainticket
  • DatasetOB.py: Dataset only for OnlineBoutique-OnlineBoutique
  • preprocessing:
  • TrainMix.py: train TEMAML in cross-system contexts using Ray
  • TrainOB.py: train TEMAML for Trainticket-Trainticket contexts
  • TrainTT.py: train TEMAML for OnlineBoutique-OnlineBoutique contexts using Ray
• imple_details: hyperparameter settings for training AttenAE and TEMAML
• requirements.txt: lists the dependencies (packages and their versions) required to run the project.

Datasets

• This project uses open datasets: Nezha (https://github.com/IntelligentDDS/Nezha) and DeepTraLog (https://fudanselab.github.io/DeepTraLog/). We do not hold the right to publish these datasets here. Please refer to the original sources for downloading the datasets.