[FEATURE] Async Reader

[roryqi]

Code of Conduct

• I agree to follow this project's Code of Conduct

Search before asking

• I have searched in the issues and found no similar issues.

Describe the feature

We synchronize to read the data. If we asynchronize to read the data. We could improve our performance.

Motivation

Improve the performance.

Describe the solution

No response

Additional context

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

[zuston]

This means we should support prefetch when reading the data, which looks similar with the tensorflow data prefetch mechanism. And this feature could improve the performance for reduce-side.

cc @advancedxy @jerqi. Could we pin this to be as the next release feature?

[zuston]

To address the issue, we will take the following steps:

1. Support async reading with a single thread to fetch data, as the current client handler is not thread-safe. [#291][part-0] feat(client): support async read  #603
2. Make the client handler thread-safe to enable multiple threads to fetch remote data. [#291][part-1] feat(client): make read handler thread safe #785
3. Introduce memory management aligned with Spark's memory management.