Amazon SageMaker AI - Amazon Timestream

Amazon Timestream for LiveAnalytics will no longer be open to new customers starting June 20, 2025. If you would like to use Amazon Timestream for LiveAnalytics, sign up prior to that date. Existing customers can continue to use the service as normal. For more information, see Amazon Timestream for LiveAnalytics availability change.

Amazon SageMaker AI

You can use Amazon SageMaker Notebooks to integrate your machine learning models with Amazon Timestream. To help you get started, we have created a sample SageMaker Notebook that processes data from Timestream. The data is inserted into Timestream from a multi-threaded Python application continuously sending data. The source code for the sample SageMaker Notebook and the sample Python application are available in GitHub.

  1. Create a database and table following the instructions described in Create a database and Create a table.

  2. Clone the GitHub repository for the multi-threaded Python sample application following the instructions from GitHub.

  3. Clone the GitHub repository for the sample Timestream SageMaker Notebook following the instructions from GitHub.

  4. Run the application for continuously ingesting data into Timestream following the instructions in the README.

  5. Follow the instructions to create an Amazon S3 bucket for Amazon SageMaker as described here.

  6. Create an Amazon SageMaker instance with latest boto3 installed: In addition to the instructions described here, follow the steps below:

    1. On the Create notebook instance page, click on Additional Configuration

    2. Click on Lifecycle configuration - optional and select Create a new lifecycle configuration

    3. On the Create lifecycle configuration wizard box, do the following:

      1. Fill in a desired name to the configuration, e.g. on-start

      2. In Start Notebook script, copy-paste the script content from Github

      3. Replace PACKAGE=scipy with PACKAGE=boto3 in the pasted script.

  7. Click on Create configuration

  8. Go to the IAM service in the AWS Management Console and find the newly created SageMaker execution role for the notebook instance.

  9. Attach the IAM policy for AmazonTimestreamFullAccess to the execution role.

    Note

    The AmazonTimestreamFullAccess IAM policy is not restricted to specific resources and is unsuitable for production use. For a production system, consider using policies that restrict access to specific resources.

  10. When the status of the notebook instance is InService, choose Open Jupyter to launch a SageMaker Notebook for the instance

  11. Upload the files timestreamquery.py and Timestream_SageMaker_Demo.ipynb into the Notebook by selecting the Upload button

  12. Choose Timestream_SageMaker_Demo.ipynb

    Note

    If you see a pop up with Kernel not found, choose conda_python3 and click Set Kernel.

  13. Modify DB_NAME, TABLE_NAME, bucket, and ENDPOINT to match the database name, table name, S3 bucket name, and region for the training models.

  14. Choose the play icon to run the individual cells

  15. When you get to the cell Leverage Timestream to find hosts with average CPU utilization across the fleet, ensure that the output returns at least 2 host names.

    Note

    If there are less than 2 host names in the output, you may need to rerun the sample Python application ingesting data into Timestream with a larger number of threads and host-scale.

  16. When you get to the cell Train a Random Cut Forest (RCF) model using the CPU utilization history, change the train_instance_type based on the resource requirements for your training job

  17. When you get to the cell Deploy the model for inference, change the instance_type based on the resource requirements for your inference job

    Note

    It may take a few minutes to train the model. When the training is complete, you will see the message Completed - Training job completed in the output of the cell.

  18. Run the cell Stop and delete the endpoint to clean up resources. You can also stop and delete the instance from the SageMaker console