Feature: make retry strategy exponential backoff multiplier configurable

[Velfi]

NOTE: I plan to allow for user-defined retry strategies in a separate PR.

Motivation and Context

aws-sdk-rust#567

Description

Updated SDK Client retry behavior to allow for a configurable backoff multiplier.

Previously, the backoff multiplier (named r in the code) was set to 2 seconds. This is not an ideal default for services like DynamoDB that expect clients to quickly retry failed request attempts. Now, users can set quicker (or slower) multipliers according to their needs.

#[tokio::main]
async fn main() -> Result<(), aws_sdk_dynamodb::Error> {
    let retry_config = aws_smithy_types::retry::RetryConfigBuilder::new()
        .max_attempts(4)
        .backoff_multiplier(Duration::from_millis(20));

    let shared_config = aws_config::from_env()
        .retry_config(retry_config)
        .load()
        .await;

    let client = aws_sdk_dynamodb::Client::new(&shared_config);

    // Given the 20ms backoff multiplier, and assuming this request fails 3 times before succeeding,
    // the first retry would take place between 0-20ms after the initial request,
    // the second retry would take place between 0-40ms after the first retry,
    // and the third retry would take place between 0-80ms after the second retry.
    let request = client
        .put_item()
        .table_name("users")
        .item("username", "Velfi")
        .item("account_type", "Developer")
        .send().await?;

    Ok(())
}

For a request that gets retried 3 times, when base is 1 and initial_backoff is 2 seconds:

• the first retry will occur after 0 to 2 seconds
• the second retry will occur after 0 to 4 seconds
• the third retry will occur after 0 to 8 seconds

For a request that gets retried 3 times, when base is 1 and initial_backoff is 30 milliseconds:

• the first retry will occur after 0 to 30 milliseconds
• the second retry will occur after 0 to 60 milliseconds
• the third retry will occur after 0 to 120 milliseconds

Testing

I wrote some test for the backoff calculation and ran existing retry tests

Checklist

• I have updated CHANGELOG.next.toml if I made changes to the smithy-rs codegen or runtime crates
• I have updated CHANGELOG.next.toml if I made changes to the AWS SDK, generated SDK code, or SDK runtime crates

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[rcoh]

this looks great! the one change I'd make is considering renaming with_backoff_multiplier to initial_backoff—it's what we call it in the interior code and I think it's a bit more grokable. But totally open to discuss this

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[rcoh]

backoff multiplier -> initial delay

[rcoh]

Nice! LGTM

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.