<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>PD 分离 on Hellokitty&#39;s Blog</title>
    <link>https://yangyang233333.github.io/tags/pd-%E5%88%86%E7%A6%BB/</link>
    <description>Recent content in PD 分离 on Hellokitty&#39;s Blog</description>
    <generator>Hugo</generator>
    <language>zh-CN</language>
    <lastBuildDate>Mon, 24 Aug 2026 16:58:43 +0800</lastBuildDate>
    <atom:link href="https://yangyang233333.github.io/tags/pd-%E5%88%86%E7%A6%BB/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>vLLM 与 Mooncake 对接源码解读：Prefill/Decode 分离中的 KV Cache 如何跨节点传输</title>
      <link>https://yangyang233333.github.io/posts/vllm-mooncake-integration/</link>
      <pubDate>Mon, 24 Aug 2026 16:58:43 +0800</pubDate>
      <guid>https://yangyang233333.github.io/posts/vllm-mooncake-integration/</guid>
      <description>深入分析 vLLM 内置 MooncakeConnector：从代理拆分请求、Scheduler 元数据、GPU KV Cache 注册，到 Mooncake Transfer Engine 通过 RDMA 将 Prefill KV 直接搬到 Decode 节点。</description>
    </item>
  </channel>
</rss>
