TianyiChen0316 / GLO

Code for paper GLO: Towards Generalized Learned Query Optimization

GLO: Towards Generalized Learned Query Optimization

Requirements

Before running GLO, please install PyTorch and Deep Graph Library (DGL) following the instructions on the pages, and run the command pip install -r requirements.txt to install the required packages. (We use torch==2.0.1 and dgl==1.1.1+cu118 in our experiments.) In addition, we use psqlparse to parse queries. Since we found that installing psqlparse with pip may result in errors, we recommend building it directly with setup.py.

Running

Please use the following command to train GLO.

python glo.py

GLO provides a list of settings as follows.

• -F FILE_ID: To change the output file ID of experiment results and checkpoints. The default value is (default).
• -d TRAIN TEST: To train with the specified training and testing workloads. TRAIN is the folder of training workload, and TEST is for testing workload.
  • The specified path is relative to ./GLO/. For example, when the training workload path is ./GLO/dataset/train_tpcds, please use dataset/train_tpcds instead.
• -e EPOCHS: To train with the specified number of epochs. The default value is 200.
• --seed VALUE: To set the random seed of training.
• -D DATABASE: To specify the PostgreSQL database name. By default, we set the name to database.
• -U USER: To specify the PostgreSQL user name. The default value is postgres.
• -P PASSWORD: To specify the password of the PostgreSQL user. By default, the password is not used.
• --port PORT: To specify the port of PostgreSQL.
• --host HOST: To specify the server host name of PostgreSQL.
• --reset: To ignore the previous checkpoint and retrain the model.
• --warm-up ITER: To execute the train and test workload queries for ITER iterations before training. This argument can be used to warm up the shared buffers.