Introduction to Practicus AI Worker Nodes
Some advanced Practicus AI features require to use software in addition to Practicus AI APP or the developer SDK. In this section we will learn how to use Practicus AI Worker Nodes.
What is a Worker Node?
Some Practicus AI features such as AutoML, making AI Predictions, Advanced Profiling and production deployment capabilities require a larger setup, so we moved these from the app to a backend (server) system.
You have multiple Worker Node options to choose from. You can run them in the cloud or on your computer. Please view help on choosing a Worker Node to learn more.
Setup Worker Node
Please check the Setup Guide to learn how to configure Practicus AI Worker Nodes.
You can use the free cloud tier for this tutorial, or use containers on your computer as well.
Launching a new Worker Node
- Click on the Cloud button to open the Worker Nodes tab
- Make sure the selected local for your computer, or the optimal AWS Cloud Region. The closest region geographically will usually give you the best internet network performance
- Click Launch New
- Pick the type (size) of your Worker Node
- Click ok to launch
The default size will be enough for most tasks. You can also choose the free cloud tier.
In a few seconds you will see your Worker Node is launching, and in 1-2 minutes you will get a message saying your Worker Node is ready.
Stopping a Worker Node
If you use local container Worker Nodes you have less to worry about stopping them.
Cloud Worker Nodes
Similar to electricity, water, or other utilities, your cloud vendor (AWS) will charge you a fee for the hours your Worker Node is running. Although Practicus AI Worker Nodes automatically shut down after 90 minutes, it would be a practical approach to shut down your Worker Nodes manually when you are done for the day.
For this, you can simply select a clod node and click on the Stop button. The next day, you cna select the stopped Worker Node, click Start and continue where you are left.
Tip: It is usually not a good idea to frequently stop / start instances. Please prefer to stop if your break is at least a few hours for optimal cost and wait time.
Terminating a Worker Node
Practicus AI Worker Nodes are designed to be disposable, also called ephemeral. You can choose a Worker Node and click Terminate to simply delete everything related to it.
Please be careful that if you choose to store data on the local disk of your Worker Node, this will also get lost after termination. In this case, you can prefer to copy your data manually, or simply click the Replicate button before terminating a Worker Node.
(Optional) Using Jupyter Lab
For technical users.
Every Worker Node comes with some external services preconfigured, such as Jupyter Lab, Mlflow, Airflow.
- Select a Worker Node that is running and ready
- Click on Jupyter button
This will start the Jupyter Lab service and view inside the app. You can also right-click tab name and select Open in browser to view the notebook on your default browser.
- If you shut down the app, the secure connection tunnel to the Worker Node notebook service will be lost even if the Worker Node continues to run.
- There are two separate Conda kernels configured for your notebook server. Big data one will have common libraries and data engines, such as DASK, RAPIDS (if you have GPUs) and Spark installed. The ML one, as the name suggests, will have ML related libraries such as scikit-learn, Xgboost, Pycaret ..
(Optional) Using the Terminal
For technical users.
You can choose a Worker Node and click the Terminal button to instantly open up the terminal. You have sudo (super-user) access and this can be a very powerful and flexible way to customize your Worker Node.