Automatically trigger sagemaker inference endpoints

If the maximum timeout of AWS Lambda is not sufficient for your video processing needs, you can explore alternative approaches to automatically invoke your SageMaker inference endpoint when new videos are uploaded to Amazon S3 or schedule it to run on batches of videos every night. Here are a couple of suggestions:

AWS Step Functions: AWS Step Functions is a serverless workflow service that enables you to coordinate multiple AWS services into serverless workflows. You can create a Step Function workflow that triggers the SageMaker inference endpoint and handles the processing of your videos. Here's a high-level overview of the approach:

Configure an event notification on your S3 bucket to trigger an AWS Lambda function whenever a new video file is uploaded. This Lambda function will start the Step Function workflow. Define a Step Function workflow that includes a task to invoke your SageMaker inference endpoint for each video file. You can use the AWS SDK or Boto3 library in a Lambda function to call the SageMaker endpoint. Handle the batching or parallel processing of videos within the Step Function workflow, ensuring that the processing time stays within the allowed limits. Configure the workflow to run automatically every night using a scheduled event (e.g., AWS CloudWatch Events). With this approach, you can handle videos of any length as the Step Function workflow can span longer durations without the limitations of a Lambda function.

AWS Batch: AWS Batch is a fully managed service for running batch computing workloads. It allows you to efficiently run large-scale, parallel, and high-performance computing (HPC) applications in the AWS Cloud. You can use AWS Batch to process your videos in batches, either as individual videos or as a group.

Here's a high-level overview of the approach:

Configure an event notification on your S3 bucket to trigger an AWS Lambda function whenever a new video file is uploaded. This Lambda function will create a job in AWS Batch. Define an AWS Batch job that runs the processing logic for a single video or a group of videos. Configure AWS Batch to handle the necessary resources and parallelism required for your video processing. You can specify the desired instance types and count to optimize the processing speed and cost. Schedule AWS Batch jobs to run every night using AWS CloudWatch Events or other scheduling mechanisms available in AWS (e.g., Amazon EventBridge). This approach allows you to process videos in batches and efficiently utilize computing resources for parallel processing.

By utilizing AWS Step Functions or AWS Batch, you can overcome the time limitations of AWS Lambda and automate the invocation of your SageMaker inference endpoint when new videos are uploaded or on a scheduled basis. Choose the approach that best suits your requirements and workload characteristics.