Rate limiting requests to Amazon AWS services

[lonix1]

I've found many examples of using Polly with HttpClient. However I'm using AWS' .NET SDK which includes a custom client.

The service I'm using (SES) limits requests to 50,000 per day and 14 per second. I'd like to run as many requests in parallel as possible, without exceeding the limit, and without queuing (if I hit the limit, there should be backpressure rather than queuing, which is computationally expensive).

Are there any simple samples of using Polly with AWS, Azure, etc.?

If not, and I must start from scratch, am I right in assuming that I should investigate the "Rate Limiting Policy"? Does it handle the above use case alone, or should it be paired with something else?

[martincostello]

Can you share a link to a specific HttpClient example you're looking at?

There are some HttpClient-specific ways to integrate Polly that "hide" it from the consumer by hooking into HttpClientFactory and adding Polly into the HTTP request pipeline transparently. The Microsoft extensions for Polly use this approach, for example.

Apart from the above, HttpClient isn't special and any code can be wrapped with Polly to do retries, circuit-breaking, rate-limiting, timeouts etc.

Our documentation for rate-limiting should be applicable for AWS, Azure or anything else.

If you want to go down the HttpClient-specific route, you would need to tie the creation of AWS clients into the HttpClient pipeline so that instances are retrieved from HttpClientFactory, rather than through its own-built-in mechanism.

For example you could create a custom implementation of Amazon.Runtime.HttpClientFactory that uses IHttpClientFactory internally, and pass this to AWSSDK clients. This would then allow any Polly functionality associated with HttpClient to be automatically applied to AWSSDK clients provided they use your adapter/bridge/wrapper.

[lonix1]

Can you share a link to a specific HttpClient example you're looking at?

I was referring to AmazonSimpleEmailServiceV2Client which seems to be a custom implementation; from what I can tell it doesn't even wrap HttpClient.

Our documentation for rate-limiting should be applicable for AWS, Azure or anything else.

That seems to do what I need. I just wanted to be sure I'm going in the right direction. Many thanks for your advice!

[martincostello]

All AWSSDK classes derived from AmazonServiceClient use HttpClient internally. They eventually come from here by default. There's just lots of hoops to jump through to stitch it together with HttpClientFactory.