Add AWS S3 Connector with Streaming support

[mfelsche]

Describe the problem you are trying to solve

It is very common in event processing to stream data to some kind of persistent storage engine for later processing or archiving purposes. One very prominent storage engine is AWS S3.

A common practice is to stream data into files that aggregate across a time window (e.g. 1 hour) or that accumulate a certain number of events or grow to a certain size. An AWS S3 connector should support this style of streaming.

Describe the solution you'd like

We would like to have an AWS S3 Connector that enables tremor to read S3 objects in a streaming fashion (source-part of the connector) and to write data to S3 objects also in a streaming fashion (sink-part of the connector).

It should support all the common ways of authentication to AWS and maintain authentication across the whole lifetime of the connector (e.g. through token refresh etc.).

It should use the official Rust SDK: https://github.com/awslabs/aws-sdk-rust

[rahul799]

Hey @mfelsche,
This project seems very interesting to me so I will like to work on this as part of the LFX'21 Mentorship program. Thank you.

[mfelsche]

Nice!

Please apply via the LFX site once it appears here as a mentorship. This might take some days: https://mentorship.lfx.linuxfoundation.org/#projects_accepting

We will handle it from there. Here is a tutorial-like guide from the LFX on how to apply: https://docs.linuxfoundation.org/lfx/mentorship/mentees/apply-to-a-project

[dak-x]

Hi @mfelsche .

• The awk-sdk in currently available as alpha and I donot see any roadmap for the official release yet. Does the support has to be experimental right now?
• smithy-rs (the code generation tool for the sdk) does not produce runtime agnostic code. So are we expected to contribute this to their sdk as well. I will have to check whether they are accepting contribution for smithy-rs. I just hope this is straightforward.

[mfelsche]

Hi @dak-x it would be cool to not rely on the tokio runtime that is used for the rust aws-sdk, but changing the codegen tool for the aws-sdk (smithy-rs) is not a requirement. It would be wicked cool, nontheless 😎

[mfelsche]

Also i wouldnt worry about the SDK being experimental. This is fine!

[bshang17]

Hi @mfelsche,

I have applied this mentorship program via LFX, this project seems interesting. I got some experience with some other Object Storage Service like Aliyun OSS. Hope I could get the opportunity to work on this project.

Thanks,

[mfelsche]

This is done