文件最后提交记录最后更新时间
12 天前
7 天前
12 天前
12 天前
README.md

Quick reference

Mooncake | openEuler

Current Mooncake docker images are built on the openEuler. This repository is free to use and exempted from per-user rate limits.

Mooncake is a high-performance transfer engine for distributed AI inference workloads, developed by KVCache.AI. It provides a RDMA/TCP-based data plane for efficient KV cache transfer between disaggregated prefill and decode instances in large language model (LLM) serving systems. Mooncake is designed for distributed storage and high-speed data movement in cloud-native AI infrastructure.

Learn more about Mooncake on GitHub⁠.

Supported tags and respective Dockerfile links

The tag of each mooncake docker image is consist of the version of mooncake and the version of basic image. The details are as follows

Tag Currently Architectures
0.3.11.post1-oe2403sp3 Mooncake 0.3.11.post1 on openEuler 24.03-LTS-SP3 amd64, arm64

Usage

In this usage, users can select the corresponding {Tag} based on their requirements.

  • Pull the openeuler/mooncake image from docker

    docker pull openeuler/mooncake:{Tag}
    
  • Run the transfer engine benchmark (sender side)

    docker run --rm --network host openeuler/mooncake:{Tag} \
        transfer_engine_bench --mode=initiator \
        --metadata_server=<metadata-ip>:2379 \
        --local_server_name=<local-ip>:12345
    
  • Run the transfer engine benchmark (receiver side)

    docker run --rm --network host openeuler/mooncake:{Tag} \
        transfer_engine_bench --mode=target \
        --metadata_server=<metadata-ip>:2379 \
        --local_server_name=<local-ip>:12346
    
  • Validate the transfer engine

    docker run --rm --network host openeuler/mooncake:{Tag} \
        transfer_engine_validator --metadata_server=<metadata-ip>:2379
    
  • View available tools

    Tool Description
    transfer_engine_bench Benchmark the transfer engine throughput and latency
    transfer_engine_validator Validate transfer correctness between two nodes
    transfer_engine_bench_with_notify Benchmark with notification-based completion

Question and answering

If you have any questions or want to use some special features, please submit an issue or a pull request on openeuler-docker-images.