Lock Prediction for Zero-Downtime Database Encryption

This repository contains supplementary material for capturing lock data from IBM Db2 and training models to predict a sequence of locks. For more information, refer to the preprint.

Setup

1. Create a Python 3.11 virtual environment: See 0_prep.sh.

   python3 -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt

   Note: For the narval system, see requirements_narval.txt.

2. Download and extract the data to the data/ directory.

• The data archive can be downloaded here.

Project Structure

• .vscode/: Holds debug configuration file specific to VS Code
• .github/workflows: Contains GitHub Actions workflows for continuous integration and automation
• analysis/: Directory for results analysis R scripts and notebooks
• data/: Stores all downloaded and extracted data for the project
• experiments/: Experiment runner scripts for different configurations (includes SLURM integration)
• logs/: Stores experiment log files
• play_scripts/: Contains utility data and modelling exploration notebooks
• results/: Directory to store experiment results
• src/: Contains the main source code
  • train.py: Main script for training models
  • datapipeline.py: Data loading and preprocessing
  • model.py: Model architectures (Transformer and LSTM)
  • evaluate.py: Evaluation metrics and result reporting
  • utils.py: Utility functions
  • tests/: Unit tests
• workload-testing/ Directory includes the HammerDB scripts to start TPC-C workloads

TPC-C Workload Simulation

All file names referenced in this subsection are relative to the workload-testing/ directory.

• Enable the DB2 trace using the command in traceCommands.txt.
• Start TPC-C HammerDB workloads by using hammerDBinit.bat, hammerDBStep1.bat, and hammerDBStep2.bat.
• Stop the DB2 trace using the command in traceCommands.txt.
• Start trace preprocessing by running db2trc flw -t.
• Extract the row and table locks using lookup_locknames.py.

Running Experiments

Several experiment scripts are provided in the experiments/ directory. To run an experiment:

bash experiments/exp-2.sh

For SLURM-based systems, the same script can be run using the SLURM command:

sbatch experiments/exp-2.sh

Note: Most experiment scripts are interopable with SLURM and normal bash execution.

Data Processing

After running experiments, use the extract and transform scripts to process the results. See the scripts for more details:

bash 2_extract.sh
bash 3_transform.sh

Testing

Run unit tests using pytest command:

pytest

Citation

If you use or study the code, please cite it as follows.

@article{rakha2025lockprediction,
  title={Lock Prediction for Zero-Downtime Database Encryption},
  author={Mohamed Sami Rakha and Adam Sorrenti and Greg Stager and Walid Rjaibi and Andriy Miranskyy},
  journal={arXiv preprint arXiv:2506.23985},
  year={2025},
  doi={10.48550/arXiv.2506.23985}
}