How to import and export data from Amazon ML to RDS?

Question

A customer is running ML algorithms on AWS, at this moment they are reading data from RDS into an Amazon Machine Learning instance and training it.

They want to know, if we have some examples on how to from the Machine Learning instance they can pick those results and then through API or SDK (preferent serverless) and feed them back into an RDS database.

I was searching for some examples but I was not successful, do anyone here have hands on experience with this or can provide me a guide on where to look?

Jonathan_T · Accepted Answer

Amazon ML will output the predictions to a specified S3 bucket. I guess the real question is more about how to effectively import data from S3 to RDS rather than an ML question.

For this, you have several options:

- Using Data Pipeline with the Load S3 Data into RDS template: http://docs.aws.amazon.com/datapipeline/latest/DeveloperGuide/dp-template-copys3tords.html
 - Use S3 Trigger that will trigger a Lambda function each time a new prediction is generated - this is the serverless approach but that means you will have as many Lambda calls as you have predictions - not necessarily the most cost-effective solution if you have a lot of predictions. The Lambda function can then handle inserting back into RDS
 - To automate this a bit further, look into CloudWatch Metrics that you could monitor (http://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/ml-metricscollected.html) to trigger the insert job

How to import and export data from Amazon ML to RDS?

Relevant content