Visitar URL original
[FEATURE] Dedicated faster serialization when shuffle writing/reading · Issue #2596 · apache/uniffle · GitHub
Skip to content

[FEATURE] Dedicated faster serialization when shuffle writing/reading #2596

Description

@zuston

Code of Conduct

Search before asking

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

Describe the feature

After overlapping compression is applied, the majority of shuffle write times is the seralization. Is it possible to replace the kyro serialization with the fast implementation like apache fory? It looks it has a impressive benchmark result .

https://github.com/apache/fory/blob/main/docs/benchmarks/serialization/bench_serialize_STRUCT2_to_array_tps.png

Motivation

No response

Describe the solution

No response

Additional context

No response

Are you willing to submit PR?

  • Yes I am willing to submit a PR!

Activity

  1. zuston commented on Aug 28, 2025

    @zuston
    MemberAuthor

    cc @jerqi

  2. roryqi commented on Aug 28, 2025

    @roryqi
    Contributor

    cc @jerqi

    There are some points that:

    1. The type system of serialization
    2. supportsRelocationOfSerializedObjects
  3. roryqi commented on Aug 28, 2025

    @roryqi
    Contributor
  4. zuston commented on Aug 28, 2025

    @zuston
    MemberAuthor

    cc @jerqi

    There are some points that:

    1. The type system of serialization
    2. supportsRelocationOfSerializedObjects

    Yes. I will draft a simple impl to benchmark

  5. added a commit that references this issue on Aug 28, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions