Glue - Predicate pushdown with Dynamodb

Question

Hello,

```
   dynamo_df = glueContext.create_dynamic_frame.from_options(
    connection_type="dynamodb",
    connection_options={"dynamodb.input.tableName": lkup_table,
        "dynamodb.throughput.read.percent": "1.0",
        "dynamodb.splits": "100"
    }
    )  
```

It seems Glue is loading entire dynamodb table (lkup_table). 
If I add filter, like  dynamo_df .filter(col('case')=='1234') - 1.Spark first loads entire table into df 2.Then it filterout the records which isn't efficient way.
Is there anyway to add predicate pushdown that avoids complete table load into dataframe (dynamo_df)?
Pl. suggest

Accepted Answer

There is for s3 tables but unfortunately not for DynamoDB.  
What you can do is minimize the performance hit (and cost) by using the new s3 export for DynamoDB.  
Check this blog: https://aws.amazon.com/blogs/big-data/accelerate-amazon-dynamodb-data-access-in-aws-glue-jobs-using-the-new-aws-glue-dynamodb-elt-connector/

Answer

Unfortunately DynamoDB does not support predicate push down syntax, as its a NoSQL database and to apply the filter the entire table would need to be read regardless.

https://docs.aws.amazon.com/glue/latest/dg/aws-glue-programming-etl-partitions.html

If this is a one-time read then you can consider the [export to S3 capability](https://aws.amazon.com/blogs/big-data/accelerate-amazon-dynamodb-data-access-in-aws-glue-jobs-using-the-new-aws-glue-dynamodb-elt-connector/) but if you intend on reading continuously you may just want to read the table directly to get more up-to-date data.

Glue - Predicate pushdown with Dynamodb

Relevanter Inhalt