OpAdviser

Create Dev Container

JetBrains Gateway/PyCharm

https://www.jetbrains.com/help/pycharm/connect-to-devcontainer.html

Visual Studio Code

https://code.visualstudio.com/docs/devcontainers/containers#_quick-start-open-a-git-repository-or-github-pr-in-an-isolated-container-volume

Create venv

JetBrains Gateway/PyCharm

https://www.jetbrains.com/help/pycharm/configuring-python-interpreter.html

Visual Studio Code

https://code.visualstudio.com/docs/python/environments

Prepare workload

bash ./workload_preparation.sh

Starting point

Run scripts/optimize.py

Below is original README.md

OpAdviser: An Efficient Transfer Learning Based Configuration Adviser for Database Tuning

OpAdviser is a customized and efficient tuning system that addresses the search space construction and the search optimizer selection problems for database configuration tuning.

Installation

Installation Requirements:

• Python >= 3.6

  git clone [email protected]:Blairruc-pku/OpAdviser.git && cd OpAdviser
  pip install -r requirements.txt
  pip install .

Preparation

Workload Preparation

Please reffer to the details instuction for preparing the workloads.

Database Connection Setup

To provide the database connection information, the users need to edit the config_auto.ini.

db = mysql
host = 127.0.0.1
port = 3306
user = root
passwd =

Quick Start

1. Specify the tuning objective in config_auto.ini. Here are some examples.

   Performance tuning, e.g., maximizing throughputs.

   task_id = op1
   performance_metric = ['tps']

   Setup automatic space construction and optimizer recommendation

   ##path of data repository
   data_repo = ../repo
   ##Turn on space construction
   space_transfer = True
   only_knob = False
   only_range = False
   ##Turn on optimizer recommendation
   auto_optimizer = True
   auto_optimizer_type = learned

2. Conduct Tuning.

   cd scripts
   python optimize.py  --config=config_auto.ini

Contact

If you have any technical questions, please submit new issues.

If you have any other questions, please contact Xinyi Zhang[[email protected]].