SageMaker Feature Store is full of questions for me, ever since I started working with it.
The thing which is absolutely not clear for me is about integration with Athena/Glue.
Since there is no built-in option in FeatureStore to, for instance, get all records, or get records by criteria, I have to use AthenaQuery branched off the FeatureGroup for that purpose.
In case if I do something like
feature_group = FeatureGroup(name='<...>', sagemaker_session=sagemaker_session)
query = feature_group.athena_query()
query.run('SELECT * FROM <whatever> LIMIT 1000', output_location='s3://...')
query.wait()
query.as_dataframe()
Will I also be charged for querying Athena? What's about Hive?
If yes, then what's the purpose of using FeatureStore? It's API is too poor for implementing continuous training, - I'm instead forced to use at least Athena, and, therefore, pay for the extra service besides I already pay for FeatureStore.
So, when I'm planning cost efforts in AWS Calculator, after calculating SageMaker Feature Store, should I also add Athena as a separate service, or it's already included in the Feature Store cost estimate?