English | 简体中文
A-Tune is an OS tuning engine powered by AI. A-Tune uses AI technologies to enable the OS to understand services, simplify IT system tuning, and maximize application performance.
Supported OS: openEuler 20.03 LTS or later
yum install -y atuneFor openEuler 20.09 or later, atune-engine is needed.
yum install -y atune-engineyum install -y golang-bin python3 perf sysstat hwloc-guiyum install -y python3-dict2xml python3-flask-restful python3-pandas python3-scikit-optimize python3-xgboost python3-pyyamlOr
pip3 install dict2xml Flask-RESTful pandas scikit-optimize xgboost scikit-learn pyyamlIf you have already installed the database application and want to store A-Tune collection and tuning data to the database, you must also install the following packages:
yum install -y python3-sqlalchemy python3-cryptographyOr
pip3 install sqlalchemy cryptographyTo use the database, you should also select either of the following methods to install dependency for the database application.
| Database | Install Using yum | Install Using pip |
|---|---|---|
| PostgreSQL | yum install -y python3-psycopg2 | pip3 install psycopg2 |
git clone https://gitee.com/openeuler/A-Tune.gitcd A-Tune
makemake collector-install
make installYou can run the following command to query the NIC that needs to be specified for data collection or optimization and change the value of the network configuration item in the /etc/atuned/atuned.cnf file to the specified NIC.
ip addrYou can run the following command to query the disk that needs to be specified for data collection or optimization and change the value of the disk configuration item in the /etc/atuned/atuned.cnf file to the specified disk.
fdisk -l | grep devsystemctl daemon-reload
systemctl start atuned
systemctl start atune-enginesystemctl status atuned
systemctl status atune-engineYou can save the newly collected data to the A-Tune/analysis/dataset directory and run the model generation tool to update the AI model in the A-Tune/analysis/models directory.
Format
python3 generate_models.py
Parameter Description
| Parameter | Description |
|---|---|
| --csv_path, -d | Path for storing CSV files required for model training. The default directory is A-Tune/analysis/dataset. |
| --model_path, -m | Path for storing the new models generated during training. The default path is A-Tune/analysis/models. |
| --select, -s | Indicates whether to generate feature models. The default value is false. |
| --search, -g | Indicates whether to enable parameter space search. The default value is false. |
Example:
python3 generate_models.py
This command is used to list the supported profiles as well as active profiles.
Format:
atune-adm list
Example:
atune-adm listThis command is used to manually activate the profile to make it in the active state.
Format:
atune-adm profile
Example: Activate the profile corresponding to the web-nginx-http-long-connection.
atune-adm profile web-nginx-http-long-connectionThis command is used to collect real-time statistics from the system to identify and automatically optimize workload types.
Format:
atune-adm analysis [OPTIONS]
Example 1: Use the default model to identify applications and perform automatic tuning.
atune-adm analysisExample 2: Use the user-defined model for recognition.
atune-adm analysis --model /usr/libexec/atuned/analysis/models/new-model.mUse the specified project file to search the dynamic space for parameters and find the optimal solution under the current environment configuration.
Format:
atune-adm tuning [OPTIONS] <PROJECT_YAML>
Example: See the A-Tune offline tuning example. Each example has a corresponding README guide.
For details about other commands, see the atune-adm help information or A-Tune User Guide.
A-Tune-UI is a web project based on A-Tune. Please check A-Tune-UI README for details.
We welcome new contributors to participate in the project, and we are happy to provide guidance for new contributors. Please sign CLA before contribution.
If you have any question, please contact A-Tune.
The SIG meeting is hold at 10:00-12:00 AM on Friday every two weeks. Please send your issues to the A-Tune mailing list.
