Skip to content

Future Updates and document Inquiry #1

Description

@ryan910707

Hello,

Thank you for sharing your work. I found the RDMA-based optimizations in your paper very interesting and would love to explore the implementation further. I am interested in running the project and understanding its details.

Could you let me know if this repository will receive future updates and provide related documents? Also, what is the current status of this repository?

Looking forward to your response!

Best regards,
yi syuan

Activity

  1. changed the title [-]Future Updates and RDMA Optimization Implementation Inquiry[/-] [+]Future Updates and document Inquiry[/+] on Mar 18, 2025
  2. cctry commented on Mar 18, 2025

    @cctry
    Contributor

    Hi,

    I am working with other authors to complete it. To run this, you can install the RDMA library at https://github.com/rainj-me/rdma-transport. Then you can run KVDirect/decode_worker.py and KVDirect/prefill_worker.py to launch the works.

    Note that this requires the IB network, and we did not test this on other hardware. The open-sourced RDMA library have some minor bugs. Please let me know if you encounter any and I will ask them to release a newer version.

  3. ryan910707 commented on Apr 1, 2025

    @ryan910707
    Author

    Hi,

    Thank you for your reply.
    I'm also very interested in the push/pull mode you mentioned in the paper. I would like to ask how you implemented these two communication methods, and which part of the code I can refer to. Also, is there a benchmark for comparison with push/pull mode?

    Thank you!

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