Nectar Applications - AI Ready

The Nectar Research Cloud has packaged AI Ready environments into Applications to support your artificial intelligence (AI) and data science research on Nectar.

When using the AI Ready applications, consider creating a persistent volume and making a GPU reservation. These optional steps should be followed before launching an AI Ready application.

Persistent volume

Prior to deploying an AI Ready application, you may create a Persistent Volume to store your user data when using the application. For more information, please read the support article on Persistent Volume Storage. Please note that the application must use the same Availability Zone (AZ) as the persistent volume.

The advantages of using persistent volume are that the volume storage size can be much larger than what is provided, and as the persistent volume is stored externally, the application can be destroyed without deleting the persistent volume.

In practice, when you launch a new AI Ready application, you can provide the same username and password, select the same persistent volume and AZ, and your data and settings on the volume will migrate to the new instance. This is important when using the GPU reservation service for AI training and inference applications. 

Follow the tutorial to create Volume Storage. Once your application has been deployed you can also follow the tutorial steps to mount the volume you created.

GPUs

For intensive workloads, you will also find it beneficial to make a GPU reservation. Attaching a GPU to your AI Ready application will provide a significant computational boost.

How to launch the AI Ready application

Step 1: Create a Persistent Volume (Optional)

Follow the tutorial to create a persistent volume. This involves visiting the Project / Volumes / Volumes menu on the Nectar Dashboard, selecting Create Volume button, and completing the fields to create a volume.

Step 2: Create a GPU Reservation (Optional)

Follow the steps under How do I access it? in the support article on the Nectar GPU/Large memory reservation service.

Step 3: Create a Key Pair

Follow the steps in the online tutorial to create a keypair for use in Nectar.

Step 4: Quick Deploy

On the Nectar Dashboard, navigate to Applications / Browse / Browse Local using the navigation panel on the left.

Find the AI Ready application, and click the Quick Deploy button. If you cannot initially see the AI Ready application then you may need to either click the Next Page button at the bottom of the page or use the Filter search bar at the top right to search for the AI Ready application.

Step 5: Configure the Application

When you click Quick Deploy, a window will open for you to Configure the Application.

5a. Python environment

Select the Python environment that best suits your activity. The available options include:

• Base Python Environment
• PyTorch
• PyTorch and TorchVision
• TensorFlow
• Generative AI and Large Language Models (LLMs)

For more information on these environments, please read our support article on Python environments for the AI Ready application (article coming soon).

Click the Next button in the bottom right to continue.

5b. Configure settings

Use the drop-down menus to select your desired Instance flavor. Note the flavors used for AI Ready images have been filtered to those that will give you better results for the demands of AI-enabled research. If you have a GPU or huge RAM reservation, relevant flavors will also appear in the list for you to select.

Add the Key Pair you created in Step 3.

When selecting an Availability zone (AZ), please ensure you use the same AZ as your persistent volume if you are recreating the instance from an existing volume.

The default Disk size is 100 GB. Please consider your workload needs when selecting an appropriate Disk size.

Click Next to continue.

5c. Host and DNS (Optional)

You can optionally choose a Host name and DNS zone for your application instance. If you choose a DNS zone, we will automatically create a DNS entry for you and provision a HTTPS security certificate for your instance. For more information, please read the support article on the Nectar DNS Service.

Click Next to continue.

5d. Configure the user account

Provide your desired Username and Password for Jupyter, R-Studio, Grafana, Airflow, and MLFlow. You will later be prompted for these once the application has been deployed.

Click Next to continue.

5e. Application Name

The last step allows you to provide an Application Name. You can keep the default or supply your own text. This may be helpful if you plan on deploying other application environments in the future.

Click the Create button in the bottom right corner of the window. 

You will be taken to the Application Environments page.

Step 6: Deploy the Environment

Navigate to Applications / Applications / Environments on the Nectar Dashboard, and click on the desired environment for more information and to deploy.

On the right side of the screen, above the created application, click the Deploy This Environment button.

The application will start deploying. Please note that it usually takes approximately 5 to 10 minutes to complete the deployment.

When the application environment is deployed and ready, the Last operation will appear with the address of your application instance. Click this link to get started. A new tab will open.

Step 7: Log into the Application

Enter the Username and Password you provided in Step 5d.

The landing page will load. You can now launch your selected application.

Step 8: Mount the Persistent Volume (Optional)

If you followed Step 1: Create a Persistent Volume (Optional), you can now follow the tutorial to mount the volume you created.