DynamoDBContext BatchGet in Parallel

[danny-zegel-zocdoc]

Describe the feature

Currently DynamoDBContext's BatchGet functionality will chunk requests with more keys than the DDB BatchGet limit into multiple requests which is great but it seems to process them in series. Ideally all the BatchGet chunks would be processed in parallel.

Use Case

Using DynamoDBContext to perform a BatchGet operation with more keys than the DDB BatchGet API allows.

Proposed Solution

When using DynamoDBContext to BatchGet with many keys such that the SDK chunks the request into multiple API calls, invoke those calls in parallel instead of in series.

Other Information

No response

Acknowledgements

• I may be able to implement this feature request
• This feature might incur a breaking change

AWS .NET SDK and/or Package version used

AWSSDK.DynamoDBv2

Targeted .NET Platform

.NET 8 and up

Operating System and version

N/A

[ashishdhingra]

@danny-zegel-zocdoc Thanks for opening the feature request.

Currently DynamoDBContext's BatchGet functionality will chunk requests with more keys than the DDB BatchGet limit into multiple requests which is great but it seems to process them in series. Ideally all the BatchGet chunks would be processed in parallel.

Could you please elaborate the feature request with example and analysis which demonstrates above existing behavior? Are you using async API as demonstrated here or sync API here? Are you instead proposing to get hold of Task object from each awaited call and then execute Task.WaitAll() to wait for all tasks to complete?

Thanks,
Ashish

[danny-zegel-zocdoc]

Could you please elaborate the feature request with example and analysis which demonstrates above existing behavior? Are you using async API as demonstrated here or sync API here?

I'm using using the async method, specifically IDynamoDBContext.CreateBatchGet<T>() and then invoking ExecuteAsync() on the resulting BatchGet<T> object.

The async code you linked appears to be the very code that shows the async batches are awaited in the while loop, i.e. in series, and not in parallel.

Are you instead proposing to get hold of Task object from each awaited call and then execute Task.WaitAll() to wait for all tasks to complete?

Yes one way of parallelizing the calls would be to collect the Task objects from each network call instead of awaiting them immediately and then using Task.WaitAll() to await them in parallel.

Thanks for considering this improvement.

[ashishdhingra]

Needs review with the team.